The Tn3-family of Replicative Transposons

Tn3-family S-resolvases are typically small (∼180 to 210 aa) comprising a relatively well conserved ∼120-aa N-terminal catalytic domain connected to a short ∼65-aa C-terminal DNA-binding domain (Fig. 6). However, the resolvases of some transposons (e.g., ISXc5 and Tn5044 from Xanthomonas campestris or Tn5063 and Tn5046 from Pseudomonas sp.) are larger (∼310 aa) due to a ∼110-aa C-terminal extension (18–21) (Fig. 6). The role of this extension is not known. Its deletion abolished Tn5044 resolvase recombination in vivo (18) but had no effect on ISXc5 resolvase activity (22). As previously observed (7), the S-resolvase phylogeny does not perfectly match that of the transposases, suggesting that cointegrate resolution modules were independently acquired or exchanged between transposons of the different subgroups after the divergence of their transposition module (Fig. 4).

The transposon resolution site, res, is generally located immediately upstream of the resolvase gene (Fig. 6). The best-characterized Tn3-family res sites, those of Tn3 and γδ (also called Tn1000), are ∼120 bp long and contain three subsites (I to III). Each subsite is composed of inversely oriented 12-bp resolvase recognition motifs (Fig. 6). Site I is the recombination core site (or crossover site) at which strand exchange occurs; sites II and III are regulatory elements required for assembly of a topologically defined synaptic complex to control recombination directionality (see Convergent mechanisms to control the selectivity of recombination section, below).

The three-subsite organization of res is conserved in most Tn3-family transposons in spite of the differences in resolvase sequence specificity. There is some variation, however, in the spacer length between the core site I and the accessory sites II and III, ranging from four to seven integral helical turns (Fig. 6). Conservation of the helical phasing between sites I and II is important to properly align the core and regulatory resolvase subunits in the synaptic complex (23, 24).

A structurally different res site organization has been proposed for other transposons such as ISXc5 in which the site II inverted repeats are replaced by direct repeats (7, 25). Tandem resolvase binding motifs were also identified in the putative res sites of several Tn3-family transposons from Gram-positive bacteria (e.g., Tn1546 and TnXO1) (7, 25). These res sites also appear more compact with two resolvase binding sites instead of three (Fig. 6). This two-subsite organization was initially described for Sin and the β-recombinases of Gram-positive plasmids to which the Tn1546 and TnXO1 resolvases are only distant relatives (25, 26) (Fig. 4 and 6). Recombination mediated by these recombinases requires the host protein Hbsu (or any DNA bending protein like the nucleoid-associated protein HU, the integration host factor IHF, or high-mobility group of proteins [HMG] from eukaryotes) that binds the spacer segment between sites I and II, playing an architectural role in synaptic complex assembly (25, 27–31). Other serine recombinases involved in plasmid monomerization are more closely related to resolvases from cointegrate resolution systems (32) (Fig. 4). In some cases there is clear evidence that they were derived from transposons during recent evolution (25, 32).

Finally, related resolution systems are also found in a distinct class of transposable elements called “Mu-like transposons” for their resemblance to bacteriophage Mu. These include β-lactam resistance transposon Tn552 from Staphylococcus sp. and transposons of the widespread Tn5053/Tn402 family (33–35) (Fig. 4 and 6). Like Tn3-family transposons, these move by replicative transposition and generate cointegrates. The Tn552 BinL resolvase is a close homolog of the plasmid pI9789 BinR recombinase, whereas the Tn5053/Tn402 TniR recombinase is closer to the resolvase of classical Tn3-family members like Tn163 and Tn5393 (Fig. 4). TniR and BinL are also distant relatives of the C-terminally extended resolvases (e.g., those of ISXc5 and Tn5044; Fig. 4) and the organization of the resL site of Tn552 was proposed to resemble the resolution site of ISXc5 (7, 25) (Fig. 6).

Y-recombinases fall into well-separated subgroups based on sequence similarities, which correlate with different recombination site organization (Fig. 5 and 7). As for the S-resolvases, recombinases of the different subgroups are phylogenetically related to other Y-recombinases that play a role in stabilizing circular replicons (Fig. 5).

Recombinases of one group, exemplified by TnpI of Bacillus thuringiensis transposons Tn4430 (285 aa) and Tn5401 (306 aa) (36, 37) are distant homologs of the chromosomally encoded multifunctional XerCD recombinases whose primary function is to resolve chromosome dimers arising from homologous recombination before their segregation into daughter cells (38–40) (see also the chapter by C. Midonet and F.-X. Barre in this volume).

The TnpS/OrfI recombinases (323 aa and 351 aa, respectively) encoded by the Pseudomonas sp. Tn4651/Tn5041 transposon subfamily (41–43) represent a second protein group more closely related to bacteriophage P1 Cre protein and to recombinases of a variety of large Rhizobiacae plasmids (9) (Fig. 5). Interestingly, TnpS-mediated or OrfI-mediated cointegrate resolution requires the product of a divergently oriented gene (TnpT or OrfQ) that shows no homology to any other characterized protein (41, 44).

Another putative Y-resolvase type identified in Tn4556 from Streptomyces fradiae is more closely related to the ResD resolvase of Escherichia coli F factor than to the two other groups (7, 45) (Fig. 5). Finally, the defective Pseudomonas putida transposon Tn4655 carries a unique site-specific recombination module composed of a recombination site (attI) and a Y-recombinase (TnpI) substantially larger (415 aa) than the other resolvases of the family (46). This system was proposed to have been initially involved in gene exchange between different replicons before incorporation as a bona fide cointegrate resolution system (46).

The IRS of Tn4430 (116 bp) is located immediately upstream of tnpI (Fig. 7). It is composed of four TnpI-binding motifs with a common 9-bp sequence. Two 16-bp inverted motifs (IR1 and IR2) form the IRS core recombination site, whereas the adjacent 14-bp direct repeats (DR1 and DR2) are dispensable but required to control selectivity and directionality (36, 47) (Fig. 7) (see also Variation on a theme: the TnpI/IRS recombination complex of Tn4430 section, below). The IRS of Tn5401 exhibits a similar arrangement of TnpI binding motifs suggesting a similar mechanism to control cointegrate resolution (37, 48) (Fig. 7).

A completely different organization has been proposed for the Tn4651/Tn5041 subfamily resolution site (designated “rst” for the resolution site associated with TnpS and TnpT) (41, 44). Deletion analysis located the minimal requirement for full rst activity to a 136-bp sequence between tnpS and tnpT (Fig. 7). DNA strand exchange takes place at a 33-bp inverted repeat (IRL-IRR), dividing rst into two arms (44). A similar 34-bp cross-over site was previously proposed for Tn5041 (42). The accessory protein, TnpT, binds to another conserved 20-bp inverted repeat element (IR1–IR2) in the tnpT-proximal arm side of rst (44) (Fig. 7). However, it has yet to be determined how this protein collaborates with TnpS to mediate recombination.

Tn3-family members that confer resistance to antimicrobial compounds are found in different subgroups of the family. In the simplest cases, these elements encode one or more determinants active against a specific class of compounds, sometimes associated with additional regulatory functions. Examples of these include the TEM-1 β-lactamase genes—Tn3 and its derivatives (1), the inducible macrolide-lincosamide-streptogramin resistance gene ermAM—Tn917 from Enterococcus faecalis (56), the aminoglycoside phosphotransefrase strA and strB genes—Tn5393 (57) or the tetracycline efflux pump tetA(A) gene—Tn1721 (58). Dissemination of vancomycin resistance by Tn1546 (10.85 kb) from Enterococcus faecium involves a cluster of seven van genes. Five of these are required for the inducible synthesis of alternative peptidoglycan precursors that are refractory to the glycopeptide antibiotic (59).

Mercury-resistance transposons, which encode genetic determinants allowing detoxification of mercuric ions from the environment, are also found in different Tn3 subfamilies. They are evolutionarily ancient and widely dispersed elements found in a variety of bacterial species and diverse environments (20, 60, 61). The mercury resistance operons (mer) comprise a variable number of genes depending on the element and its detoxification capabilities. In most cases, they include a metal-responsive regulator (merR), a transport system responsible for delivering mercuric ions into the cell (merTP), and a cytoplasmic mercuric reductase (merA) that converts toxic Hg²⁺ ions into less toxic reduced forms (61).

With the successive development of novel antibiotherapies, transposons initially active against only restricted ranges of compounds, some of which like mercury resistance transposons predated the antibiotic era, accumulated additional resistance genes, as the consequence of coselective pressure (see below). There are now many identified Tn3 family representatives with different combinations of antibiotic resistance determinants, which contribute to the emergence and accelerated expansion of multi-resistant bacterial pathogens in both the clinical and community environments (3, 4, 61, 62).

Tn3-family catabolic transposons are typically large (>40 kb) and enable the host cells to use xenobiotic compounds such as recalcitrant aromatic hydrocarbons (e.g., toluene, xylenes, and naphthalene) as a sole source of carbon and energy. This generally relies on complex clusters of genes that often comprise an “upper” pathway operon that is responsible for the conversion of the initial substrate into oxidized intermediates, a “meta” operon that is required to degrade the resulting intermediates into central metabolites, and additional regulatory genes that are involved in the transcriptional activation of the whole pathway (5, 6).

Other functions associated with Tn3-family transposons include virulence factors from both animal and plant pathogens. For example, the B. anthracis virulence plasmid pXO1 transposon TnXO1 (8.7 kb) carries the gerXBAC operon responsible for spore germination during infection (63, 64) and the Pseudomonas syringae TnhopX1 promotes dispersal of HopX1, a conserved effector protein present in most plant pathogens with type III secretion systems (65). Recent bioinformatics analyses reveal that the contribution of Tn3-related elements to plant virulence is more general, being involved in the spread of different classes of secreted type III effectors in the Xanthomonads (A. Varani and M. Chandler, personal communication). A few transposons from separate subgroups encode post-segregational killing toxin–antitoxin systems (59, 66, 67). This presumably helps to stabilize the transposon in a bacterial population.

Acquisition of new phenotypic traits may occur when the proximal end of a transposon inserted adjacent to a selectable marker gene is inactivated and a more distant “surrogate” IR sequence is recruited. This was suggested to occur during acquisition of the carbapenemase bla_KPC gene by the transposon Tn4401 (68) and presumably in other single-trait Tn3-family transposons. In Tn3-related elements, a degenerate “ancestral” IR sequence is found immediately upstream of the bla_TEM1 passenger gene (our unpublished observation); whereas in Tn4401 the supernumerary internal IR end was inactivated by insertion of ISKpn7, which also reinforced expression of the bla_KPC gene (68–70).

Unitary transposons may also be derived from composite transposons by inactivation of one of the two flanking transposon copies. This is the case for the tetracycline resistance transposon Tn1721, in which one element of the initial composite structure is inactivated by deletion leaving the other element (Tn1722) intact and capable of autonomous transposition (58, 71) (see also Fig. 2). In addition, Tn1721, like several other Tn3-family members, mediates one-ended transposition, a poorly understood transposition reaction catalyzed by the transposase at a single IR end that generates recombinant products containing one terminus of the transposon and variable lengths of adjacent DNA (72–76). This can directly provide single-end derivatives for the construction of new modular arrangements without requiring subsequent deletion events.

Combination of functional transposition modules is well illustrated by certain catabolic transposons that form large (up to ∼90 kb) mosaic structures containing multiple mobile subentities (e.g., (46, 77–79)). This is viewed as an adaptive strategy allowing movement and reassortment of different combinations of catabolic functions within and between separate plasmids from environmental bacteria (6). A similar nested organization has been reported for other phenotypically distinct Tn3-family transposons such as the Yersinia enterocolitica lactose Tn951 (80) or the mercury-resistance derivatives Tn510 and Tn511 (81). Some of these are transposition-defective and require an active copy of a related element to be mobilized in trans (46, 77, 80, 81).

There are specific recombination systems that facilitate acquisition of new passenger genes. A particularly striking example is the recruitment of mobile integrons, site-specific recombination systems that use a Y-recombinase family integrase (IntI) to promote the exchange of specific gene cassettes (82–84) (see also the chapter by J. A. Escudero et al. in this volume). This led to the rapid and incremental diversification of antibiotic resistance genes in the Tn21-related mercury-resistance transposons and other Tn3-family members (85, 86). The evolution of multidrug-resistant transposons has been proposed to retrace the recent history of antimicrobial treatments (for discussion, see references 2, 3, 87, and 88).

A proposed founding event at the onset of multidrug resistance evolution was the incorporation of a specific integron and its associated cassettes into a “Mu-like” Tn5053/Tn402-family transposon (33), generating a specific class of “mobile” integrons known as “class 1” or “Class 1 In/Tn” (82, 87). Although this may have produced different variants of the ancestral Class 1 In/Tn (89, 90), the key to introduction and spread of one particular type of Class 1 integron among bacterial pathogens was presumably the presence of a gene (qacE) conferring resistance to quaternary ammonium compounds, which were commonly used as biocides before the antibiotic era. Concurrently, bacteria also accumulated mercury-resistance transposons, notably because of industrial enrichment of Hg²⁺ in the environment and the use of mercury-containing disinfectants in clinical practices (2, 3, 88). Since Tn402 family transposons preferentially integrate within res sites (91–93), coselection of both types of mobile element could have led to the formation of recombinant transposons with Class 1 In/Tns inserted within or close to the res site of mercury-resistant Tn3-family members. Integration of different variants occurred not only at separate positions within different Tn21-subgroup members but also in non-mercury-resistance transposons such as Tn1721 (85, 86). With the introduction of successive classes of therapeutic antibiotics, the integrons of these different transposons subsequently acted as a platform for collecting and redistributing new resistance cassettes from the reservoir, thereby accelerating their accumulation and dissemination among both established and emerging pathogens.

Passenger genes can also be incorporated into Tn3-family members in the form of composite transposons (e.g., TnHad2, (77)) or through the promiscuous transposition activity of members of certain IS families (e.g., ISEcp1 [IS1380 family] or ISCRs [IS91 family]) able to transduce variable lengths of DNA adjacent to their integration site (2, 85, 94–97). ISEcp1 and ISCR1 are for example responsible for incorporation of variants of extended-spectrum bla_CTX-M genes into Tn21-related elements (e.g., Tn1722), or into class 1 integrons with which they are associated (98–102).

Another major process driving Tn3 family evolution is recombination between both related and unrelated transposons (85). This may occur by homologous recombination between common regions or by inter-transposon site-specific recombination at the res site (6, 20, 61, 85). Resolvase-mediated recombination between heterologous res sites has been inferred from sequence comparison (1, 20, 79, 103) or demonstrated experimentally (78, 104, 105). Res site recombination can occur both intra- or inter-molecularly and can also lead to inversions between adjacent transposons reflecting some promiscuity in the activity of certain S-resolvases, which are normally controlled by a mechanism of resolution selectivity (78, 105, 106) (see Convergent mechanisms to control the selectivity of recombination section, below). Similar activities were proposed for the TnpS/OrfI Y-recombinase of the Tn4651/Tn5041 subgroup (43, 44) and for the atypical Tn4656 TnpI/attI recombination system (46). Promiscuous recombination mediated by cointegrate resolution systems appears to be an ongoing process that constantly reshapes the Tn3 family by exchanging functional module and passenger genes between its different members (6, 20, 61, 85).

The most conserved transposase region overlaps with a ∼27-kDa proteolytic fragment roughly corresponding to the C-terminal third of the protein and contains 9 of the 15 perfectly conserved residues including the three acidic residues that constitute the “DD-E” catalytic triad (7, 11) (Fig. 9). This motif is the signature of a superfamily of proteins, the DD-E/D polynucletotidyl transferases, that includes various nucleic-acid processing enzymes, as well as different transposase families and the retroviral integrases (121–123). These enzymes have a common “RNase H-like” fold of their catalytic core, in which the conserved triad provides the essential active site residues. Phosphodiester bond breakage and resealing reactions uses an Mg²⁺-dependent transesterification mechanism with a water molecule or a 3′ OH group as a nucleophile (123, 124). The Tn3-family DD-E motif is surrounded by other highly conserved amino acid residues underlining the importance of this region. Mutation of the conserved Asp751 residue in the Tn4430 transposase DD-E motif abolishes DNA cleavage and strand transfer in vitro, whereas the other activities (i.e., ability to bind and pair the transposon ends) are unaffected (Nicolas et al., in preparation).

Secondary structure predictions suggest that the DD-E region exhibits the βββαβαβ-α topological arrangement typical of the RNase H fold, with the conserved catalytic residues occupying the expected positions within the predicted structure (120). This was confirmed by three-dimensional modeling using different protein threading tools, with the best fits being obtained with the catalytic core of HIV-1 integrase (Fig. 9) (M. Lambin and B. Hallet, unpublished data). However, the proposed Tn3-family RNAse H fold domain is predicted to contain an extra ∼90-aa insertion between the last β strand of the central β sheet and the C-terminal α helix carrying the conserved Glu residue of the DD-E catalytic triad (120) (Fig. 9). This insert corresponds to one of the most conserved regions of the proteins, suggesting an important role in transposase activity. Insertion domains with different structures are found in an equivalent topological location of the Tn5 and Hermes transposases where they are proposed to play a role in the formation of DNA hairpin intermediates for the excision of these elements (122, 125–127).

Another highly conserved Tn3-family transposase segment lies at the N-terminus. Structural homology searches based on secondary structure predictions and fold recognition modeling revealed that this region has a high likelihood to form a bipartite DNA-binding domain analogous to that of the human centromeric protein CENP-B (128) (Fig. 9). As for the C-terminal RNase H fold, this prediction was verified for a consensus sequence derived from the different Tn3-family transposase subgroups, reflecting the high level of sequence similarity in this region (120) (Fig. 9). Modeling is consistent with previous results that identified the IR-binding domain at the N-terminus of the transposase (129, 130). Deletion of this region abolishes end binding by Tn4430 TnpA in vivo and in vitro, while a 300-aa N-terminal fragment of the protein binds the IR end with the same affinity as the full-length transposase (E. Nicolas and B. Hallet, unpublished data).

The CENP-B DNA-binding domain consists of two helix-turn-helix motifs (HTH1 and HTH2) separated by a short, 10-aa linker (Fig. 9). Both motifs contact adjacent major DNA grooves making specific interactions with the target sequence. The N-terminal residues (residues 1 to 9) form a flexible arm that makes additional contacts with the phosphate backbone (128). These different structural components of the CENP-B binding domain appear to be conserved in the N-terminal domain of Tn3-family transposases based on secondary structure prediction and sequence comparison (120) (Fig. 9).

Intriguingly, the CENP-B DNA-binding domain is structurally homologous to that of Tc1/mariner transposon family and to the bipartite end-binding domain of bacteriophage Mu transposase, MuA (131–133). CENP-B is believed to have evolved from an ancestral transposase that has been “domesticated” during evolution to acquire a novel function (134). Confirming the presence of a CENP-B-like DNA-binding domain at the N-terminus of Tn3-family transposases would therefore provide an additional example illustrating the modular organization of the DD-E/D transposases, which presumably resulted from the recruitment of separate functional domains early during evolution (122, 123).

In the crystal structure of the Mos1 transposase (a Tc1/mariner-family member) and MuA, the N-terminal DNA-binding domain is immediately fused to the RNase H fold domain by a short connector of a few amino-acid residues (131, 133). In transposons of the Tc1/mariner family, this linker region contributes to the transposase dimer interface and plays a role in reciprocal communication between transposase subunits during transposition (131, 135, 136). In Tn3-family transposases, the equivalent region of the protein is occupied by a large central domain, corresponding to a ∼63-kDa proteolytic fragment of Tn4430 TnpA (residue 152 to 682) (Fig. 9). This domain is predicted to have an all-alpha structure. Despite several patches of well-conserved residues, it shows no sequence or structural homology with other proteins (120). However, filter binding assays performed with the Tn3 transposase suggested that separate subregions of the central domain (residue 243 to 283 and residue 439 to 505) exhibit nonspecific DNA-binding activity (130).

A possible function of the unique central TnpA domain might be in target immunity. Most transposons and ISs do not exhibit this property and this would explain why Tn3-family transposases are significantly larger than other transposases. A genetic screen was developed to identify Tn4430 TnpA mutants that could bypass immunity by promoting high levels of transposition into an immune target (i.e., a target carrying a Tn4430 IR end). The selected T⁺/I⁻ (i.e., transposition-plus/Immunity-minus) mutations were clustered into five discrete regions belonging to separate domains. Although two clusters of mutations were identified within the large central domain, those that showed the strongest defect in immunity mapped in the C-terminus, beyond the DD-E motif. This scattering of the T⁺/I⁻ mutations indicates that target immunity is likely to be a complex process involving multiple activities of the transposase (120) (see also The mechanism of target immunity: ‘schizophrenic-like’ behavior of the transposase, below).

As for other transposable elements, the ∼38-bp terminal IRs of Tn3-family transposons are composed of two functional domains (Fig. 10). The outermost highly conserved 5′-GGGG-3′ motif is essential for transposition, but is not required for stable TnpA binding. It contains the DNA cleavage site and is thought to be contacted by the transposase catalytic domain to carry out the chemical steps of transposition (137–142). TnpA binding to this region generally gives weak protection as determined by DNA footprinting, with some enhanced signals due to protein-induced DNA distortions around the cleavage site (140, 141, 143–148). For Tn4430, these signals are more pronounced upon formation of a paired-end complex, which corresponds to an activated intermediate of the transposition reaction (Nicolas et al., in preparation).

The inner segment of the IR (position ∼10 to 38) is the recognition domain with which the transposase makes the strongest and most specific interactions. This region also encodes the specific determinants to confer target immunity (138–140, 149, 150) (see also below). As shown in the alignment (Fig. 10), it can be further divided into two subdomains; a central AT-rich region (box B1, positions ∼11 to 27) and a more conserved subdomain (box B2, position ∼31 to 38) carrying the 5′-TAAG-3′ motif at the inward-facing boundary of the IR (Fig. 10). TnpA makes extended interactions with both recognition subdomains, protecting up to three adjacent major grooves (137, 141, 149). Mutations in the innermost box B2 subdomain showed the strongest defects in transposase binding and transposition activity (139, 140).

This binding pattern is consistent with the predicted CENP-B-like bipartite Tn3-family transposase DNA-binding domain (Fig. 9). In Tc1/Mariner-family transposases and MuA, a structurally equivalent domain recognizes sequence elements of about the same length (∼22 to 26 bp) as Tn3-family transposases (∼27 bp). Their HTH motifs interact with both sides of the recognition element making specific contacts within the major groove (131–133, 151). The basic linker connecting both HTHs forms a structure termed ‘AT-hook’ that makes additional interactions with the AT-rich stretch that separates both HTH binding sites (131, 132). It is tempting to propose that Tn3-family transposases bind the transposon ends in a similar way, with one predicted HTH motif interacting with box B2 of the IR and the other, together with the linker, interacting with the longer AT-rich box B1. Additional protein segments, such as the conserved N-terminal arm of the DNA-binding domain may also contribute to the interaction, as observed in the CENP–B/DNA complex (128) (see also Fig. 9).

A number of Tn3-family transposons have significantly longer terminal inverted repeats than the canonical ∼38-bp IR (18, 44, 48, 65, 146, 148, 152). For some of them, this was found to correlate with the presence of additional binding sites for accessory proteins in the subterminal regions of the transposon, providing a means to control transposition at the level of transpososome assembly.

The inner segment of the 53-bp Tn5401 terminal inverted repeats includes an extra 12-bp binding motif for the TnpI recombinase (37, 48, 148). This motif is not included in the IRS and is therefore not required for cointegrate resolution (48). By contrast, TnpI binding to this motif is essential for the formation of a stable complex between the transposase and the transposon ends (148). At the same time, as was previously reported for the resolvase of other Tn3-family members (7), TnpI binding to the IRS down-regulates the expression of both tnpI and tnpA, keeping transposition activity at a low level (148). This antagonistic action of TnpI may be important to ensure efficient transposition under certain circumstances. In particular, bursts of TnpI and TnpA expression would occur when the transposon enters a new host cell, before transcriptional repression by TnpI becomes effective. Functional coupling between TnpI and TnpA activities may also be regarded as a safeguard mechanism to maintain the integrity of both the transposition and cointegrate resolution modules as the transposon multiplies and spreads.

Transposase binding to the ends of γδ and Tn4652 is influenced by the host protein IHF (integration host factor). Since the cellular abundance of nucleoid-associated proteins like IHF varies according to the growth rate (153, 154), their contribution to transposition provides a means of communicating environment signals or the physiological state of the cells to the transpososome.

For γδ, transposase and IHF binding to the transposon ends is cooperative (145). This was found to primarily impact on target immunity (155) and to rescue transposition defects resulting from specific ends mutations (139). For Tn4652, IHF binding is a prerequisite for transposase binding and transposition (146). IHF was also found to up-regulate Tn4652 transposition by stimulating the transcription of the tnpA gene (156). TnpA expression is also under strong dependence of the stationary-phase sigma factor σ^s (RpoS) (157). Together, these different mechanisms co-operate to restrict Tn4652 transposition under carbon starvation conditions, when the cellular concentration of IHF is maximal and σ^s is induced. Consistently, binding of Tn4652 transposase to the transposon ends is negatively affected by Fis (factor for inversion stimulation), which unlike IHF is maximally expressed in actively growing cells (147). Fis binding to its cognate sites is proposed to out-compete IHF from the left end of Tn4652, thereby compromising the recruitment of the transposase. Transposition of Tn4652 is also modulated by the Pseudomonas putida ColRS two-component system (158) and the small transposon-encoded protein TnpC, which appears to down-regulate the expression of the transposase at the post-transcriptional level (159).

If the requirement for specific host factors is important to correlate the rate of transposition with cell physiology, it may also restrict the host-range of transposons as was proposed for IS1071 (152). This element is characterized by unusually long terminal IRs (110 bp) containing potential binding sites for host accessory proteins such as IHF (152).

Tn3-family transposons are characterized by a low level of target specificity, which likely contributed to their wide dissemination. Early studies indicated that Tn3-related elements preferentially insert into specific AT-rich DNA regions (161–164). This was confirmed from large-scale analyses which revealed that the target sites of different family members conform to a similar, low-specificity consensus sequence centered on a degenerated 5-bp AT-rich core motif corresponding to the target duplications (i.e., 5′-TA[A/T]TA-3′ for Tn3, 5′-TATAA-3′ for Tn917, 5′-[T/A] [T/A]N[A/T] [A/T]-3′ for TnhopX1, 5′-T [T/A] [T/A] [A/T] [A/T]-3′ for Tn4652, and 5′-[T/A]NNN[A/T]-3′, for Tn4430) immediately flanked by GC-rich positions (65, 165–168) (E. Nicolas and B. Hallet, unpublished data).The overall palindromic organization of this consensus likely reflects symmetrical interactions made by the transposition complex to transfer both transposon ends to the target DNA. G/C to A/T transition steps at the DNA cleavage sites may be required to distort or bend the DNA, making the target scissile phosphates more accessible for strand transfer as proposed for other transposable elements (169–171).

In all analyses, the target DNA sequence is clearly not sufficient to determine Tn3-family insertion specificity. Specific DNA sequences perfectly matching the consensus are poor targets, whereas other sites are targeted much more frequently than expected from their resemblance to the consensus (165, 166, 172). This suggests that integration of Tn3-family transposons obeys some other structural or dynamic properties of DNA.

Recent work performed with Tn4430 indicates that relatively subtle changes made in a target molecule may profoundly alter the insertion profile (173). In particular, integration of repeated copies of a pseudo-palindromic DNA sequence (i.e., three or five copies of the E. coli lacO operator) created strong insertion hot spots between the repeats and the unidirectional replication origin of the target plasmid (173) (E. Nicolas and B. Hallet, unpublished data). This altered regional preference did not change target sequence specificity, but was dependent on the direction of DNA replication with respect to the lacO repeats (E. Nicolas, C. Oger and B. Hallet, unpublished data). A plausible interpretation is that the palindromes interfere with replication forks progression in the hot spot region (e.g., through the formation of secondary structures in the DNA), thereby provoking the accumulation of preferred DNA target structures. Consistent with this, several Tn3-family transposons preferentially insert within or close to the replication origin of target plasmids (161, 174). On the other hand, Tn917 transposition into the chromosome of Enterococcus faecalis and other Firmicutes showed a strong preference for the replication termination region (172, 175, 176). This region is subject to extensive processing, including numerous replication forks collapse and repair events (177)), which may provide the required transposition signal. Intriguingly, the Tn4430 transposase binds in vitro with high affinity to branched DNA substrates that mimic a replication fork (E. Nicolas, C. Oger and B. Hallet, unpublished data).

Transposition of other transposons is also coupled to replication (178–185). Tn7 targets the lagging strand at a replication fork (178, 179, 185) using specific interactions between the Tn7-encoded auxiliary protein TnsE and the β-clamp processivity factor of the replisome (186) (see also the chapter by J. E. Peters in this volume). This results in a marked preference for inserting into conjugating plasmids and is regarded as a mechanism to promote horizontal dispersal. For other transposable elements such as the IS200/IS608 family and Group II introns, coupling between DNA targeting and lagging-strand DNA synthesis is inherent to the transposition mechanism that requires single-stranded DNA for integration (183, 187).

The possibility that replicative transposons like those of the Tn3 family preferentially integrate into replication forks would provide a direct mechanism for recruiting the host replication machinery for the replicative transposition step. Indeed, experiments performed with conditionally replicating DNA molecules have shown that target replication is essential for Tn1/Tn3 transposition (188). Transposition into nonreplicating target DNA was essentially undetectable, whereas blocking replication of the donor molecule had no effect on transposition activity. Tn3-family transposons also prefer to insert into episomes such as phages and plasmids rather than into the chromosome, presumably because the replication dynamics of these different replicons is different (189, 190).

Target immunity was initially described for Tn3 family members (194, 195) and subsequently for other transposable elements like Mu (196, 197) and Tn7 (198, 199). In all cases, a single transposon end or a subdomain of the transposon terminus capable of binding the transposase is sufficient to confer target immunity (196, 197, 199, 200), but the presence of multiple transposase binding sites usually strengthens the phenomenon (120, 192, 201, 202). For Tn3-family transposons, the DNA determinants of target immunity coincide with the internal ∼27-bp to 30-bp recognition motif (box B) of the terminal IR (120, 138–140, 149, 150) (Fig. 10).

Experiments with Tn4430 have shown that simply attracting the transposase to a nonspecific target molecule by fusing the protein to a heterologous DNA-binding domain is insufficient to make this molecule refractory to transposition (173). This suggests that establishment of immunity requires the formation of a specific complex between the transposase and a cognate site on the target (173).

The requirement for specific interactions between the transposase and the transposon ends is only part of the self-recognition mechanism that makes each transposon of the family immunized against itself, but not against other members of the family (195, 203). This was shown in “mixing” experiments where Tn4430 and Tn1 were introduced into the same cells; an immune target for Tn4430 was not immune against Tn1 and reciprocally, a target that was immunized by Tn1 transposase was permissive for Tn4430 (120). The transposase of Tn3 family transposons therefore plays a dual role in the specificity of immunity by both communicating the immunity signal to the target and by interpreting this signal.

For Mu and Tn7, target immunity results from the interplay between the transposase (MuA and TnsAB, respectively) and an additional element-encoded protein (MuB and TnsC) that function as a molecular “match-maker” between the transpososome and the target (117, 118, 204–206) (see also the chapters by R. M. Harshey and J.E. Peters in this volume). Both proteins are DNA-dependent ATPases that bind to DNA in their ATP-bound form. By interacting with their partner transposase, MuB and TnsC stimulate transpososome assembly and DNA strand transfer (200, 207–210). As a result, MuB- or TnsC-bound DNA molecules are preferred substrates for Mu and Tn7 transposition, respectively (196, 200, 208, 211–213). Interaction with the transposase also stimulates ATP hydrolysis by MuB and TnsC, which triggers their dissociation from DNA and promotes their relocation to remote sites. As this preferentially takes place on DNA molecules that carry the end-bound MuA/TnsB proteins, displacement of MuB/TnsC makes those molecules less reactive, and hence immune against further integration (196, 200, 202, 214, 215).

For Tn3-family transposons, the transposase is the only element-encoded protein that takes part in the immunity process. However, this does not rule out the possibility that some additional cellular function or host protein contributes by modulating transposase action in the same way for all family members despite their specific sequence and self-recognition properties. The action of transposase on the target may either be “dissuasive”, by selectively restricting access to immune DNA molecules, or “attractive”, by directing the transposon toward permissive targets as is proposed for the MuB and TnsC target adaptors. Depending on their global effect on transposition, the Tn4430 TnpA T⁺/I⁻ mutations were proposed to either impair the ability of the transposase to establish immunity, or to bypass the immunity signal (whether positive or negative) by promoting efficient transposition into nonpermissive targets (120). Supporting the latter scenario, recent biochemical data obtained with selected T⁺/I⁻ TnpA mutants indicate that they have promiscuous activities, suggesting a possible link between target immunity and transposase activation (Nicolas et al., in preparation) (see also Catalysis of the transposition reactions, below).

Transposition immunity is traditionally presented as a strategy to favor intracellular and intercellular dispersal of transposable elements. For example, without immunity, Tn7 would accumulate into its high-affinity chromosome site attTn7, precluding it from spreading among bacterial populations by horizontal transfer (204, 205). Another proposed role for transposition immunity is to avoid self-destruction by autointegration (i.e., integration into itself). This is particularly crucial for bacteriophage Mu, which multiplies its genome (38.5 kb) by undergoing multiple rounds of replicative transposition in a short period of time (216). Mu target immunity is effective over ∼20–25 kb on either side of the genome ends (191, 217). This would be sufficient to protect Mu against self-integration while fully exploiting the available target DNA landscape for replication.

A serious drawback of the copy-in mechanism of replicative transposition is that the transposon remains tethered to the donor molecule through its 5′ ends after cleavage at the 3′ ends (Fig. 8). This could strongly restrict the ability of the element to reach a distant target within the cell. As a consequence, replicative transposons should exhibit a high propensity to integrate into close DNA regions belonging to the initial donor molecule. However, intramolecular transposition by the copy-in mechanism gives rise uniquely to inversions or deletions of adjacent DNA sequences, which can be detrimental for the host replicon (163, 218, 219). Therefore, the primary role of replicative transposon immunity would be to protect the host genome against such deleterious consequences of intramolecular transposition. The Tn4430 T⁺/I⁻ TnpA mutants provide a unique opportunity to test this possibility (120). Preliminary results indicate that these mutants mediate intramolecular transposition at a higher frequency than the wild-type transposase, correlating with their immunity defect (E. Nicolas, C. Oger and B. Hallet, unpublished data). It would be interesting to see whether this also correlates with a loss of fitness resulting from transposition-induced damages in the host replicon.

S-recombinases cleave all four DNA strands concurrently (Fig. 11A). This generates double-strand breaks with a two-nucleotide 3′ OH overhang on one side of the break, and a recessed end to which the recombinase is attached via a 5′-phosphoseryl bond on the opposite side. The DNA strands are then exchanged by a 180° right-handed rotation of one pair of half-sites relative to the other. Base pairing between the exchanged overhangs is necessary to properly realign the DNA duplexes and orient the DNA ends for the ligation step (Fig. 11A). This step is the reversal of the cleavage reaction, with each 3′ OH group of the cleaved ends attacking the phosphoseryl bonds in the partner (7, 9, 17, 31, 223, 224).

There has been some debate in the literature concerning the molecular mechanism of rotation (7, 225–229). The model that currently conciliates most biochemical and structural data (the “subunit rotation” model) involves a complete 180° rotation of one half of the cleaved complex relative to the opposite half (227, 230). The crystal structure of an activated form of the γδ resolvase tetramer trapped in the act of strand exchange shows that both sides of the recombinase tetramer are held together by a flat hydrophobic interface that would make such a rotational motion between recombinase subunits thermodynamically favorable without causing complete dissociation of the complex (230, 231) (Fig. 11A).

In contrast to S-recombinases, Y-recombinases sequentially exchange one pair of DNA strands at a time via the formation of a Holliday junction intermediate (Fig. 12A). Each strand exchange is a concerted process in which the 3′ phosphotyrosyl DNA–protein bond generated by DNA cleavage is subsequently attacked by the 5′ OH end of the partner strand. Exchange of both pairs of DNA strands implies that separate pairs of recombinase molecules in the complex are sequentially activated (Fig. 12A). Studies performed with different Y-recombinase family members have provided a detailed picture showing how this could be achieved at the molecular level (17, 232–234).

According to the model, the recombinase-bound recombination core sites are brought together in an antiparallel alignment, forming a synaptic complex with a pseudo-four-fold symmetry (Fig. 12A). The DNA is bent to expose one specific DNA strand from each duplex toward the center of the synapse for the first strand exchange. After cleavage, DNA strands are exchanged by extruding three or four nucleotides from the core central region, and by reannealing them to the partner strand to orient the cleaved 5′ OH ends for ligation. As for the S-recombinases, this “homology-testing” step ensures that appropriate DNA strands are exchanged during the reaction. The complex then isomerizes to activate the other pair of recombinase subunits for the second strand exchange (Fig. 12A).

For all Y-recombinases studied in detail, the activity of the core recombination complex is regulated by a cyclic network of allosteric interactions that places each recombinase subunit under the control of its two neighbors, so that diagonally opposed active sites of the tetramer are sequentially and reciprocally switched on and off during recombination (235–238). Isomerization of the tetramer involves relatively limited readjustments of DNA and proteins in the complex, and the mechanisms that promote these conformational changes to provide directionality to the strand exchange reaction appear to vary among different recombination systems (239).

Efficient S-resolvase-mediated recombination of Tn3 and γδ only takes place if two full-length res sites are present in the appropriate head-to-tail orientation on a supercoiled DNA molecule. Resolvase binding to the three res subsites (Fig. 6) promotes assembly of a multi-subunit protein–DNA complex termed a synaptosome in which the two res sites are plectonemically interwrapped, trapping three negative supercoils from the initial DNA substrate (7, 17, 227) (Fig. 11B). The specific topology of this complex is proposed to act as a topological filter to ensure that the recombination sites are in the correct configuration (240). DNA wrapping around the synaptosome is facilitated by negative supercoiling. In contrast, formation of such a complex between inversely oriented res sites, or between sites on separate DNA molecules would be topologically hindered (7, 17, 227, 240).

Synaptosome assembly and activation is a dynamic process during which selective interactions between adjacent resolvase dimers result in a succession of conformational transitions that progressively drive the complex towards its final, recombination-competent configuration (17). Reciprocal interactions between resolvase dimers bound to accessory sites II and III of both res sites are responsible for the establishment of the initial inter-wrapped structure in which three negative nodes are trapped (241, 242). Formation of this presynaptic complex is the key of the topological filter that dictates whether recombination will proceed by aligning the core recombination sites I for strand exchange. Productive parallel pairing of the crossover sites I involves specific interactions between the resolvase molecules bound to site I and site III, as well as additional interactions between the site I-bound dimers of both duplexes to form a synaptic interface within the tetramer (230, 231, 243, 244).

In addition to promoting spatial juxtaposition of the crossover sites, transient interactions between the regulatory and catalytic subunits of the complex are thought to play a more direct role in the directionality of strand exchange by inducing the required conformational change to bring resolvase from an inactive to an active configuration observed in structural studies (17, 31, 224, 230, 244). This activation step may be important to tip the recombination complex to an irreversible stage of the recombination reaction.

Because of the defined topology of the recombination complex and the rotational specificity of the strand exchange mechanism, recombination mediated by S-resolvases exclusively yields catenated molecules in which the two recombinant products are singly interlinked (Fig. 11B). The same product topology has been reported for other resolution systems using S-recombinases suggesting a common topological structure of the synaptic complex even if the organization of the recombination sites and the molecular architecture of the synaptosome are different (19, 25, 27, 30, 31) (see also the chapter by P. A. Rice in this volume).

A mechanism of toplogical selectivity analogous to that described for the serine recombinases controls cointegrate resolution mediated by the Tn4430 Y-recombinase, TnpI (47). As for S-resolvases, TnpI mediates recombination without additional host factors, acting both as a catalytic and as a regulatory component of the recombination complex. However, as opposed to most regulated site-specific recombinases, TnpI does not absolutely require the DR1 and DR2 accessory binding motifs of the resolution site IRS to be active (47) (see also Fig. 7). In the absence of DR1 and DR2, TnpI-mediated recombination at the IR1–IR2 core site is nonselective (or “unconstrained”) giving rise to all possible DNA rearrangements in vivo (i.e., deletions, inversions, or intermolecular fusions) and to topologically complex products in vitro (47). In contrast, the DR1 and DR2 accessory motifs stimulate intramolecular recombination between directly repeated IRSs and generate exclusively two-node catenane products in vitro (47) (Fig. 12B).

More recent topological analyses indicate that TnpI binding to DR1 and DR2 results in the formation of a complex in which the accessory sequences are inter-wrapped approximately three times. As for the S-resolvases, formation of this complex acts as a checkpoint (i.e., topological filter) to ensure that the recombination sites are in a proper head-to-tail configuration. Changing the arrangement between the core site and the accessory motifs inhibits recombination by compromising productive core site pairing or forcing alignment in an incorrect configuration, while mutational inactivation of DR1 and DR2 increases the level of topologically unconstrained recombination arising from random collision of the recombination sites (47, 245) (C. Galloy, D. Dandoy and B. Hallet, unpublished data).

The formation of two-node catenanes as the unique products of recombination also implies that the topologically constrained synapse between the TnpI-bound DR1–DR2 sequences imposes a specific alignment of the IR1–IR2 core sites to carry out DNA strand exchange. This alignment is such that antiparallel pairing of the core sites introduces a positive twist in the synaptic complex, thereby compensating for one negative node trapped by the accessory sequences (Fig. 12B). Specific positioning of the core sites within the recombination complex correlates with a defined order of activation of the catalytic TnpI subunits within the tetramer (245). The most distant core subunits with respect to the accessory sites (i.e., the IR1-bound subunits in the wild-type IRS) initiate recombination by catalyzing the first strand exchange, while the proximal subunits (i.e., the IR2-bound subunits) resolve the Holliday junction intermediate by exchanging the second pair of strands (245) (Fig. 12B).

According to current models, directionality of strand exchange depends on the bending direction of the core sites as they are assembled in the synapse. There are two possible recombinase tetramer configurations in which diagonally opposed recombinase subunits are activated for catalysis (Fig. 12B). If one configuration is used to initiate recombination, then the opposite configuration of the complex will terminate the reaction following isomerization (245). In the TnpI/IRS system, the choice of starting recombination with one configuration of the synapse instead of the other is primarily dictated by the DR1 and DR2 accessory motifs. In their absence, unconstrained recombination catalyzed by TnpI at the IR1–IR2 core site takes place with both possible synapse configurations and no preferred order of strand exchange (245).

Based on structural data reported for other tyrosine recombinases, positioning of the TnpI core tetramer in the synaptic complex would orient the recombinase C-terminal domains toward the regulatory region of the complex, while the N-terminal domains of the TnpI core subunits would point away from the accessory sequences (Fig. 12B). Selecting for this specific arrangement of the core complex may occur indirectly by imparting a specific path to the core sites, or more directly by requiring specific interactions between the core and accessory TnpI subunits of the complex. Imposing a specific pairing of the core recombinase complex to start recombination is important to avoid the formation of unproductive (i.e., parallel) synapses (245).

In addition to providing an architectural scaffold for assembly of the topologically and functionally selective synapse, the DR1–DR2 accessory sequences of Tn4430 IRS also affect recombination directionality by acting at later steps in strand exchange (245). Their correct orientation in the recombination substrate stabilizes DNA cleavage and rejoining intermediates that could not be observed in reactions using the minimal IR1–IR2 core site alone (245). Cleavage of the first pair of DNA strands is proposed to release the free energy stored in the topologically constrained synapse to bring about conformational changes required to generate the Holliday junction intermediate and to promote its subsequent isomerization (245). This allosteric activity of the accessory components of the complex ensures efficient substrate conversion to products by preventing reversal of the reaction.

A topologically defined complex with a different molecular architecture promotes selectivity during resolution of plasmid multimers by the tyrosine recombinase XerCD from E. coli (39, 246). In this case, assembly of the topologically constrained synapse requires dedicated cellular proteins instead of extra recombinase molecules and the recombination product is a four-node catenane instead of a two-node catenane. It is presently unclear whether similar mechanisms of selectivity function in the TnpS–TnpT/rst (OrfQ-OrfI/att5041) cointegrate resolution system encoded by the Tn4651/Tn5041 subgroup of Tn3-family transposons (41, 43, 44) (Fig. 7). Recombination mediated by the tyrosine recombinase TnpS at the rst resolution site of Tn4651 is a relatively slow process that requires the accessory protein TnpT (44). However, it is not known whether TnpT binding to rst results in the formation of a topologically constrained synaptic complex as for other resolution systems. Likewise, further biochemical studies are required to decipher the molecular mechanisms that control recombination mediated by the PmrA-like tyrosine recombinase encoded by Tn4556 (7) and by the TnpI/attI recombination system of Tn4655 (247).