Xer Site Specific Recombination: Double and Single Recombinase Systems

Plasmid Resolution: Multicopy Plasmid ColE1 and Accessory Proteins

Plasmid dimerization and eventual multimerization has been termed as the “Dimer catastrophe” due to its deleterious effect in cell populations (Summers et al., 1993; Field and Summers, 2011). Dimer catastrophe represents two major problems in bacteria; (1) unequal plasmid distribution among populations, in particular, multicopy plasmids that are more vulnerable to plasmid loss and (2) metabolic burden caused by the rapid accumulation of dimers into the host (Field and Summers, 2011; Million-Weaver and Camps, 2014). As mentioned previously, dimer resolution was originally elucidated in ColE1, resulting in the first functional characterization of XerC and subsequent identification of XerD by sequence homology to XerC (Blakely et al., 1993). These discoveries constituted a new approach for site-specific recombinases and their role in dimer resolution. Subsequent investigations led to the identification of SSR enzymes involved in dimer resolution in other plasmids of E. coli, and other bacteria. Current estimates have identified more than 1300 tyrosine recombinases where many of them are associated with other host proteins to regulate their activity, directionality, or processivity (Meinke et al., 2016). Large plasmids usually carry their own recombinase machineries adjacent to the recombination site. Whereas, small plasmids, like those in the ColE1 family, use the chromosomally encoded dimer resolution system of their host (Sengupta and Austin, 2011; Crozat et al., 2014). For ColE1 resolution, XerC/D proteins act on a specific site called cer (Figure 3B), a non-codifying region of 280 bp where two additional proteins act with XerCD to catalyze SSR reactions: the arginine repressor (ArgR) (an arginine-dependent DNA binding protein originally called XerA) (Stirling et al., 1988a,b), and aminopeptidase A (PepA) (a bifunctional transcriptional regulatory protein that reacts to environmental signals, which was originally called XerB) (Stirling et al., 1989). SSR in cer is catalyzed by XerC within a sequence of 30 bp composed of two 11 bp half sides and a central region of 8 bp. XerC and XerD bind to the left and right halves cooperatively and respectively. Strand exchanges are catalyzed by XerC to form a HJ intermediate that is eventually resolved by an uncharacterized cellular HJ resolvase to generate a recombinant product (Colloms et al., 1996; Cornet et al., 1997). The cer site is comprised of a 30 bp core recombination site and two accessory DNA sequences of ∼180 bp in length, in which one or two hexamers of PepA and one hexamer of ArgR control the reaction (Colloms, 2013). The two accessory proteins are necessary for recombination, since in their absence, plasmid dimer resolution cannot be completed. However, at abnormal high concentrations of PepA, recombination in vitro can proceed without the help of ArgR (Reijns et al., 2005; Sénéchal et al., 2010). This is also seen at the recombination site psi of plasmid pSC101 which requires XerC, XerD and PepA but not ArgR (psi dimer resolution requires another accessory protein called ArcA instead of ArgR, and the cleavage reaction is performed by XerC and XerD) (Cornet et al., 1994; Colloms et al., 1996). In cer, PepA and ArgR control recombination directionality so that dimers can only be converted into monomers and not the opposite reaction. Dimer resolution directionality caused by these two proteins involves the formation of two directly repeated cer sequences positioned in an antiparallel direction; this conformation is favored by negative supercoiling where the cer sequences are interwrapped three times around the proteins resulting in the formation of a right-handed synapse structure that brings the XerCD binding sites together. Sites in an inverted repeat position prevent right-handed formation; this ensures only dimer resolution occurs. Thus, XerC and XerD bind to the 30 bp cer synapse region and may interact with the N-terminal domains of the PepA hexamers. Whereas the ArgR protein, which is flanked by one or two PepA hexamers, might be involved in bending the DNA, tightening it and activating cer SSR by possible interaction with the C-terminal region of the Xer recombinases. Another possible function is to bring the two cer sites together and to allow PepA loading to form a nucleoprotein complex that promotes XerCD binding and recombination (Reijns et al., 2005; Sénéchal et al., 2010; Colloms, 2013). Additionally, cer also encodes for a 70 nt RNA fragment called Rcd that is only transcribed during dimer formation by the Pcer promoter and regulated in a sequence-specific manner by the FIS protein (Blaby and Summers, 2009). Rcd binds to the enzyme tryptophanase and induces a quiescent state by increasing indole production within the cell. The quiescent state permits the cell to arrest cell division and chromosomal replication but still be active metabolically. This process is thought to be part of a dimer formation checkpoint that allows the XerCD/cer system to resolve dimer formation during this pause (Field and Summers, 2011).

The Bacillus subtilis Model and the Effect of Two Translocases

The capacity to perform SSR to resolve chromosome dimers is highly distributed among bacteria and archaea. Thus, homologs of XerC and XerD have been sequenced in a variety of species (Wang et al., 2013; Crozat et al., 2014). In B. subtilis, two homologs of XerC and XerD called CodV and RipX perform dimer resolution at a 28 bp dif site (Bsdif) close to the terminus region (Figure 3D). The Bsdif region is comprised of two 11 bp half-sites with imperfect dyad symmetry where CodV and RipX bind simultaneously and a 6 bp central region where DNA exchange occurs. Both CodV and RipX share a 37 and 44% identity with the XerC and XerD respectively, and 39% between them (Sciochetti et al., 1999). CodV binds preferentially to the left half-site and preferentially cleaves the top strand whereas RipX is able to bind to both sides with preferential binding to the right-half-site and preferential cleavage of the bottom strand. Cleavage by CodV is more efficient than cleavage by RipX, which suggests that CodV performs the first strand cleavage followed by RipX in in vitro experiments (Sciochetti et al., 2001). Sciochetti et al. (1999) also demonstrated that RipX could interact effectively with the E. coli dif site, unlike CodV which showed a weaker interaction with this substrate. However, addition of XerC to RipX/dif_E.coli or XerD to CodV/dif_{E. coli} generated larger complex formation in gel retardation analysis, demonstrating protein-protein interactions between these four proteins, which confirms some conserved features of tyrosine recombinases among bacteria. This is supported by the fact that the right half-site presents highly conserved features with respect to other dif sites among some bacteria, whereas the left-half site is less conserved, which could explain why RipX can binddif_{E. coli} (Sciochetti et al., 2001). In contrast to E. coli, the synaptic complex can be brought together by the action of two DNA translocases: the membrane-associated SpoIIIE protein (Stage III Sporulation Protein E) and the soluble SftA protein (Septum-associatedFtsK-likeTranslocase of DNA). Both translocases harbor AAA+-ATPase and C-terminal domains with 56% of sequence similarity between them. SftA exhibits 50% identity with respect to the E. coli FtsK_γ domain whereas SpoIIIE exhibits a 50% of similarity to the FtsK_αβ subdomain and 42% of similarity to the FtsK_γ subdomain of E. coli (Barre, 2007). The N-terminal domains of these proteins are more divergent; SpoIIIE and FtsK share 36% of identity to respect to the four transmembrane helix whereas SftA lacks the transmembrane spanning domain (Wu, 2009; Kaimer et al., 2011). N-terminal domain variations coincide with their different location and activation in the genome. FtsK and SpoIIIE share a similar mechanism to anchor to the inner membrane of the dividing cell by their N-terminal regions; this transmembrane interaction is possibly reinforced by interactions with FtsZ or other cell division components such as FtsA or ZapA (Dubarry et al., 2010). During vegetative growth, SpoIIIE shows two predominant states: a static phase, where SpoIIIE is assembled close to future sites of cellular septation, and a mobile phase, where SpoIIIE does not occupy a specific position. Once cellular division begins, the static phase takes place when SpoIIIE is recruited by FtsZ and other division machinery proteins and is escorted to the center of the division septum (Fiche et al., 2013). SpoIIIE remains in the invaginating septum and hexamerizes independently of the cell division stages (Vegetative, division and sporulation stages) and independently of DNA interaction (Cattoni et al., 2014), suggesting that SpoIIIE assembly may not be restricted to the presence of impaired DNA and on the contrary, may be involved in normal DNA segregation as demonstrated by (Fiche et al., 2013). Experiments using high-resolution microscopy revealed that under formation of asymmetric (sporulation) or symmetric (vegetative growth) septa, the SpoIIIE concentration increased 2.5-fold around the constricting septa, even without evident formation of the septa, indicating close interaction with other components of the division machinery, that in turn regulates its activity under specific conditions (Fiche et al., 2013). SftA can be localized either to the cell center or more frequently, to the forming division septum. Although SftA lacks an integral membrane domain, the FtsZ ring recruits the enzyme and attaches it to the division septum during the initiation of cellular division, which explains its localization through the cell cycle (Kaimer et al., 2009, 2011; Kaimer and Graumann, 2011). These patterns of localization suggest that both translocases (SftA and SpoIIIE) are present at the septum at various times of segregation and that they perform DNA migration independently of each other, although SftA is only involved in DNA cytokinesis in contrast to SpoIIIE that may be involved in cytokinesis and cell division processes (Kaimer et al., 2009). DNA translocation is initially carried out by SftA during septation, probably by recognition of the 8-nucleotide SRS motifs (SpoIIIERecognition Sequences) which are similar to the E. coli KOPS sequences. The SRS motifs are mostly located on the leading strand (up to 85%), and direct translocation toward the Bsdif site (Besprozvannaya and Burton, 2014). Therefore, the primary function of SftA consists of moving chromosomal DNA until the ter regions are positioned at midcell and the origin regions migrate to each pole of the cells. The SftA may also be required for other proteins involved in cytokinesis and FtsZ positioning (Biller and Burkholder, 2009). SpoIIIE, the second translocase in B. subtilis, may take over DNA translocation working synergistically but not interchangeably with SftA, it can also function as a DNA segregation checkpoint preventing membrane fusion until chromosome segregation is completed (Kaimer et al., 2011; Fiche et al., 2013). Cattoni et al. (2014) suggested that SpoIIIE binds non-specifically to the DNA in a pre-formed hexameric open ring conformation and then searches for SRS motifs without hydrolysis of ATP. Similarly to FtsK proteins, SRS recognition by the SpoIIIE_y domain triggers allosteric modifications that activate the ATPase activity of SpoIIIE_αβ and therefore, DNA translocation (Besprozvannaya et al., 2013). Once it encounters SRS motifs, the hexameric ring changes to the closed and active form pumping the chromosome in an oriented manner by recognizing further SRS motifs and translocating it toward Bsdif (Cattoni et al., 2014). As mentioned before, SpoIIIE is actively expressed in all growing cells and is essential during sporulation to translocate the remaining DNA from the mother cell into the forespore compartment, and during vegetative growth to guarantee that concatenate formation or disrupted genomes will not affect normal cellular division. Moreover, SpoIIIE is also required for septal membrane fusion after completion of chromosome translocation. During sporulation, asymmetric septation encloses the DNA and traps 25–30% of one chromosome into the forespore. SpoIIIE pumps the remaining 70–75% by an analogous mechanism used by FtsK; the reaction only takes 20 min demonstrating its incredible speed (Demarre et al., 2013; Bose et al., 2016). The mechanism of SpoIIIE DNA translocation through the membrane is still unclear, since recent single molecule-imaging experiments still provide valid information for two main models; the paired DNA conducting channel model (Burton et al., 2010; Yen Shin et al., 2015) and the aqueous channel model (Fiche et al., 2013). This system of both translocases has been also detected in Staphylococcus aureus termed FtsK and SpoIIIE because of their amino acid homology to SpoIIIE and FtsK from B. subtilis and E. coli respectively. However, in contrast to B. subtilis system, in S. aureus both enzymes seems to present a redundant, although independent role in DNA segregation. Individual deletions of either FtsK or SpoIIIE did not exhibit major changes in chromosome segregation for S. aureus, however, when combined together they represented a major threat for S. aureus genome stability (Veiga and Pinho, 2016).

Consistent with their different roles, SftA and SpoIIIE do not colocalize during vegetative-replicative stages or sporulation. Thus, SftA in concert with FtsZ and division proteins moves chromosomal DNA away from the closing division septum. Then, upon septum closure, entrapped DNA is translocated through the SpoIIIE pore or channels into the correct compartment (either a forespore or a daughter cell). However, unlike FtsK that activates XerCD recombination reactions, neither SftA nor SpoIIIE directly activate CodV or RipX recombinases. In this case, SftA and SpoIIIE affect the CodV/RipX reaction by proper positioning of the ter region, but there is no evidence of direct interaction between these enzymes to date (Biller and Burkholder, 2009; Kaimer et al., 2011).

Multichromosome Bacteria and IMEX

Vibrio cholerae, as well as 10% of sequenced bacteria to date, possess a very distinct property among bacteria; it harbors more than one chromosome (Jha et al., 2012). One ancestral chromosome I (chrI) of 2.96 Mbp and one plasmid-derived chromosome II (chrII) or ‘chromid’ of 1.072 Mbp, encode 2,775 and 1,115 ORFs, respectively. ChrI contains most of the housekeeping genes whereas chrII contains essential genes specialized in adaptation to new environments or pathogenicity (Xu et al., 2003; Harrison et al., 2010; Kirkup et al., 2010; Val et al., 2016). Harboring two or more chromosomes have shown to be highly heritable among these bacteria, which suggests that multiple chromosomes offer a positive selective pressure to maintain them. One possible explanation is that multiple chromosomes might offer an advantageous feature against dimer formation. Val et al. (2008) showed that dimer formation increases exponentially in relation to the size of the replicons, thus, dividing a single replicon into two or more replicons may reduce this topological problem. However, genome size might not be relevant for the presence or absence of Xer/dif recombination machinery. Some large chromosomes do not require Xer/dif recombination machinery as in some Legionellales (genome size ranging from 2 to 5 Mb) whereas some small-sized chromosomes still require Xer/dif recombination machinery as demonstrated by some Rickettsiales (genome ranging from 0.85 to 1.52 Mb in size) (Carnoy and Roten, 2009).

Homologs of XerC/XerD and FtsK have been characterized on chrI, referred as XerC_{V C} and XerD_{V C} with 53 and 68% of amino acid similarity to E. coli XerC and XerD, respectively (Huber and Waldor, 2002; McLeod and Waldor, 2004). Whereas chrII does not encode any Xer recombinase involved in dimer resolution. dif-like sequences are present in both chromosomes (dif1 and dif2) located near GC skew shift-points (Val et al., 2008; Kono et al., 2011). Interestingly, both dif sites differ from each other in their sequences, dif2 harbors five different nucleotides compared to dif1 and most α-proteobacterial dif sites, four of them in the central region, resembling dif-like plasmid composition (Kono et al., 2011). Dimer resolution in V. cholerae requires FtsK_{V C} translocation by recognition of KOPS-like motifs (GGGNAGGG) in a similar way to that found in E. coli. Once the dif sites are brought together nearby, FtsK_{V C} activates XerD_{V C}, which is positioned to cleave the bottom strand, and perform the first strand cleavage. Then XerC_{V C} cleaves the top strand and performs the second strand cleavage; these reactions are carried out on both chromosomes at their respective dif sites (Figure 3E) (Val et al., 2008). Additional studies demonstrated that E. coli FtsK was able to activate 50% of the XerCD_{V C} synaptic complexes at dif1 whereas only 20% of XerCD_{V C} were activated at dif2, suggesting that the dif2 recombination process requires more accurate interactions between the FtsK proteins and the XerCD complex (Val et al., 2008). An additional feature of multiple chromosomes is their capacity to synchronize replication termination at the same time despite their different sizes (Val et al., 2016). This capacity may confer an additional regulatory control against dimer formation due to the time-lapse between the replicated chromosomes and cellular division. Demarre et al. (2014) showed that terII sites (chrII) separate earlier than terI and that this early separation keeps terII sites at midcell by the macro domain MatP/matS organization system. This restriction during concatenation induces several collisions at midcell between terII sites, increasing the number of recombinational events and the likelihood of dimer resolution. It also ensures that ter sites of bacterial chromosomes remain exclusively in mid-cell to be processed by FtsK.

Although XerC and XerD recombinases normally perform dimer resolution, they are also exploited by other replicons such as plasmids, bacteriophages, and other integrative elements. Indeed, initial studies on plasmid stability in ColE1 and phage integration of bacteriophage λ led to the discovery of XerC and the mechanistic insights of the tyrosine family (Meinke et al., 2016). In V. cholerae, the causative agent of the potentially fatal human disease cholera, XerC_{V C} and XerD_{V C} are hijacked by some vibriophages to integrate their genomes into the chromosome. They are usually referred to as IMEX (IntegrativeMobileElements ExploitingXer), and the best known ones are VGJϕ (Vibrio Guillermo Javier filamentous phage), TLCϕ (Toxic Linked Cryptic), and CTXϕ (Cholera Toxin Phage). CTXϕ is a lysogenic [(+)ssDNA] filamentous bacteriophage that encodes the A-B type enterotoxin CT in V. cholerae (Das, 2014). These three vibriophages harbor a particular attachment site (attP), a dif-like site that serves to classify the three different groups of IMEX, (CTXϕ-type, VGJϕ-type and TLCϕ-type) (Das et al., 2013). Although the components to integrate their genomes are very similar, their mechanisms of integration differ from one to the other and from their host strains. Direct ssDNA integration by CTXϕ-type phages is characterized by the formation of a ∼150 bp folded structure created by the intra-strand base pairing interaction between two palindromic attP sites (attP1 and attP2) separated by 90 nt on the ssDNA sequence (Figure 3F) (Das, 2014). The two overlap regions attP1 and attP2 reassemble the XerC_{V C} side of dif1 and dif2 regions but differ from the XerD-side. This lack of homology between XerD_{V C} recognition site and attP_CTXϕ limits the catalytic reaction to XerC_{V C} that catalyzes the complete reaction. An additional host factor called EndoIII participates in the directionality of the reaction, which blocks further rounds of strand cleavage by XerC_{V C} causing its dissociation and therefore preventing CTXϕ excision (Bischerour et al., 2012). Although XerD_{V C} is not involved in the catalytic reaction, it is still necessary for a successful integration, probably by its role in synaptic complex formation (Val et al., 2005). Once the integration is completed, host DNA replication proteins resolve the formed HJ intermediate and convert it to dsDNA. Prophage CTXϕ cannot be excised from its host since it loses the capacity to fold itself, which in turn prevents further base-pairing interactions between the attP sites, which ultimately abolishes the XerC_{V C} catalytic reaction (Das, 2014). Interestingly, CTXϕ integration in El Tor strains is only found in chrI, and it is generally associated with two other vibriophages, TLCϕ and RS1 that enable CTXϕ integration in V. cholerae genome by reconstituting a functional dif site, and by promoting CTXϕ replication and transmission (Hassan et al., 2010). In the classical biotype strains, CTXϕ usually targets both chromosomes (Faruque and Mekalanos, 2012).

Similarly, to CTXϕ, VGJϕ integration uses the XerC_{V C} catalytic reaction at the dif1 site, but unlike CTXϕ, it only harbors one dif-like attachment site (attP_{V GJ}ϕ) of 29 bp that allows its integration into the chromosome as a dsDNA. The attP central region contains four different nucleotides close to the XerD binding side with respect to the central region of the dif1 site (Figure 3G). The lack of homology at the XerD_{V C} central region side prevents XerD_{V C} participation in the catalytic reaction. Once integrated, prophage VGJϕ acquires two attP sites (attPL and attPR), equally functional for the XerC_{V C} excision reaction, in contrast to CTXϕ, where Xer recombinases can process VGJϕ excision from the host genome (Das et al., 2013). TLCϕ also depends on host encoded Xer recombinases for its integration. Its attP_TLCϕ site possesses high homology with the XerC_{V C} binding side and central region of dif1 whereas it is highly divergent from the XerD_VC binding site (Figure 3H). The prophage form of TLCϕ is almost always linked to CTXϕ integration confirming the regular synergistic interactions found in most IMEX. Paradoxically, despite the lack of homology between the XerD binding sites of dif1 and attP_TLCϕ, TLCϕ integration/excision is mediated by XerD_{V C} and then completed by XerC_{V C} resembling dimer resolution in bacteria, but independently of FtsK participation (Midonet et al., 2014).

IMEX are recombination platforms that permit bacteria to evolve and adapt through the acquisition and reordering of relevant genes. They have strengthened bacterial evolution, playing an important role in the rise of multidrug resistance, gene transfer mechanisms and virulence factors among clinically relevant bacteria (Fournes et al., 2016; Midonet and Barre, 2016). Besides the vibriophages just described above, some other relevant IMEX have been found; the gonococcal genomic island (GGI) related to pathogenic Neisseria species (Domínguez et al., 2011) and the EludIMEX-1 found in Enterobacter ludwigii (Antonelli et al., 2015). GGI is an unusually long IMEX (57 kb long) found in almost 80% of Neisseria gonorrheae strains and is involved in the expression of type IV secretion system (T4SS) genes (Christie et al., 2014). GGI carries a degenerate dif site called dif_GGI of 28 bp with a XerC-binding site and a central region homologous to the conserved Neisseria dif site (dif_Ng) and a divergent XerD-binding site. GGI insertion into the Neisseria genome follows a CDR-like process where FtsK activates XerD to perform the first strand cleavage between dif_Ng and dif_GGI followed by isomerisation of the synaptic complex and activation of XerC to perform the second strand cleavage, creating a GGI integrated form with two active Xer binding sites. Interestingly, the GGI synapse has given important clues about how IMEX might remain integrated in the host genome despite the presence of dif sites. Fournes et al. (2016) revealed by experiments in vitro that a trimeric form of the E. coli FtsK protein (t-FtsKαβγ_Ec) was unable to activate XerCD recombination at one of the two dif sites (the dif_GGI site), in fact, the XerCD/dif_GGI complex was unable to stop t-FtsKαβγ_EC translocation. As a consequence, the XerCD complex is dissociated from dif_GGI and the excision process is inhibited.

EludIMEX-1 is a 29.1-kb IMEX found in E. ludwigii (ECAA-01) that carries the bla_NMC-A gene that encodes for a serine carbapenemase. It was first characterized by Antonelli et al. (2015) when they sequenced the whole genome of a NMC-A-positive isolate of E. ludwigii. The results indicated the presence of a new 29-kb region with lower GC content when compared to the bacterial genome, indicating a possible gene transfer acquisition (Wu et al., 2012). Further analysis revealed that this region is flanked by putative XerC/XerD recombination sites with high homology at the XerC-binding site. They also determined that EludIMEX-1 insertion site in the genome was the same for two distinct species of the E. cloacae complex suggesting a possible acquisition via a XerC/XerD dependent recombination event at a specific dif-like site (Antonelli et al., 2015). Understanding of IMEX control and excision processes will provide us a better idea of how counteract the acquisition of antibiotic resistance genes in pathogenic microorganisms.

The difSL/XerS Model

The E. coli pathway of dimer resolution has been found to be highly conserved among bacteria with circular chromosomes. It was initially demonstrated by Recchia and Sherratt (1999) when they analyzed 16 eubacterial and five archaeal genomes for XerCD-CodV/RipX homologs. They showed that most eubacterial genomes possess two putative Xer recombinases whereas Archaea presented a single recombinase in three of the five genomes analyzed (Recchia and Sherratt, 1999). Subsequently, Carnoy and Roten (2009) demonstrated by doing an exhaustive computational analysis of 234 chromosomes from 156 proteobacterial species, that 87.8% of the genomes analyzed presented XerCD-like and dif-related sequences. Moreover, Kono et al. (2011) predicted by a recursive hidden Markov model method (including XerCD orthologs) that 578 out of 592 bacterial genomes with a single chromosome and 63 out of 66 genomes with multiple chromosomes presented a dif-like sequence. Additionally, they remarked how XerC and XerD are conserved in almost 60–70% of bacterial species, and 85% in proteobacterial species (Debowski et al., 2012). These results among many others led to the general view that the E. coli pathway is predominant for dimer resolution. However, dimer resolution machinery or regulation of strand exchange may differ: some processes may require or disregard accessory proteins, others may or may not require activation by translocases, some will be mediated by a XerC-first strand exchange whereas others by XerD-first strand exchange and others may need two recombinases or only one. Among these divergences and unique characteristics for each bacteria to solve dimer formation, the less studied ones are the unconventional single recombinases.

Recchia and Sherratt (1999) first mentioned the presence of single recombinases from the identification of two eubacterial genomes harboring only one Xer homolog. It was later confirmed when Le Bourgeois et al. (2007) demonstrated that some species of Lactococcus and Streptococcus use an alternative Xer recombination machinery. This new Xer complex is based on a single tyrosine recombinase called XerS (356 aa) that acts on an atypical 31 bp recombination site called difSL in the presence of dimers. Unlike E. coli, the xerS gene is found immediately adjacent to the recombination site difSL acting as a single module. The difSL site differs from most dif sites because of its large central region of 11 bp as opposed to the normally found 6–8 bp in all other dif regions (Figure 3I) (Leroux et al., 2011). Thus, difSL consists of two imperfect inverted repeat sites of different sizes separated by the central region where DNA exchange occurs. The inverted repeat region is one nucleotide longer in difSL and contains an extra nucleotide in the middle of the right inverted repeat (TTTTCTTGAAA) versus the left part of the sequence (TTTCCGAAAA). This additional spacing suggests XerS/difSL may be biased to favor binding in one-half site over the other. It was later confirmed by Leroux et al. (2011), where they also showed that XerS presented stronger interaction with the left-half site of difSL than the right-half site, and a preference for initiating the recombination reaction on the bottom strand of the difSL site. These results indicate that, although the difSL site is relatively symmetric and XerS is a single tyrosine recombinase, there is a bias for where the proteins initially bind to difSL and where they initiate the strand cleavage reaction. Thus, the left-bound monomer could activate the right-bound monomer by bending the DNA or changing the conformation of the second monomer which could explain the preferential cleavage and exchange of the bottom strand. This behavior resembles what XerC displays with weak binding but stronger strand exchange when compared to XerD (Nolivos et al., 2010; Leroux et al., 2011). This intrinsic bias alone cannot control the preference of the directionality of the strand cleavage reaction. The achievement of proper control requires the action of a SpoIIIE-like homolog translocase called FtsK^SL, a protein of 758 aa in length in Streptococcus mutans or 816 aa in S. agalactiae with low similarity at the N-terminal region between them. This low similarity does not affect its binding preference to the division septum commonly found in most proteins of the FtsK-HerA superfamily (Le Bourgeois et al., 2007). The C-terminal domain of FtsK^SL shows 41% similarity at the amino acid level in relation to FtsKE. coli with four of the five amino acids similar (QR-GN motif) involved in XerD interaction (Keller et al., 2016). On the other hand, FtsK^SL is unable to read E. coli KOPS motifs as demonstrated by Nolivos et al. (2012), probably due to the lack of common skewed octamers sequences called Architecture Imparting Sequences (AIMS) in Firmicutes, which means that KOPS sequences in Firmicutes are not as conserved as in proteobacteria (Hendrickson and Lawrence, 2006). This would also explain the divergence between FtsK_γ domains even among Firmicutes. Additionally, AIMS found in Lactococcus lactis differ in both in length and in sequence from traditional KOPS/SRS motifs, being A-rich heptamer motifs instead of the GC-rich octamer motifs (Nolivos et al., 2012). XerS also lacks critical residues found in XerD to interact with FtsK (residues RQ-QQ). Interestingly, XerS/difSL recombination occurs almost in a similar fashion to that of E. coli. Both Xer systems require FtsK_N localization at the division septum and FtsK_C translocation to achieved Xer dimer resolution. Additionally, XerS/difSL proved to be functional in E. coli, despite the lack of homology in their FtsK proteins. Further analyses on FtsK-Xer interactions are required since the exact mode of action is still speculative (Nolivos et al., 2010).

The Helicobacter and Campylobacter (difH/XerH) Model

Studies in Helicobacter sp and Campylobacter sp led to the discovery of another type of single recombinase called XerH that acts on a recombination site called difH in a FtsK-dependent manner. It was shown to be involved in chromosome segregation and possibly dimer resolution in Helicobacter pylori (Debowski et al., 2012). XerH (354–362 aa) differs from the traditional XerCD (298 aa) recombinases by its size and protein homology (26% of identity with respect to XerCD). It also shows more similarity to XerS (356 aa) in both the size of the protein and the high degree of homology of their recombination sites (Carnoy and Roten, 2009; Leroux et al., 2013). Another characteristic of XerH and a possible hallmark of single recombinases (XerS and XerH) is that the difH sequence is also located near the recombinase-encoding gene, indicating a possible individual genetic module for Xer expression (Le Bourgeois et al., 2007; Carnoy and Roten, 2009). Interestingly, most of the epsilon species of 𝜀-proteobacteria harbor a XerH/difH system whereas some other 𝜀-proteobacteria (Sulfurimonas denitrificans and Sulfurovum) possess a system analogous to the classical XerCD system. Additionally, unlike other tyrosine Xer recombinases, XerH activity appears to be affected by a second Xer recombinase called XerT in H. pylori (the TnPZ transposon associated recombinase) since under XerT deletion, difH recombination levels increased (Debowski et al., 2012). Recent structural studies showed that the Helicobacter difH comprises two highly conserved imperfect inverted binding sites of 11 and 10 bp (AGTTATGAAAA and AAAAGTTTGA) in the left and right sides respectively, separated by a 6–10 bp central region (Figure 3J) (Bebel et al., 2016) (Unpublished data suggest a 10 bp central region, Leroux et al., 2013). Two subunits of XerH bind cooperatively to each side with a stronger binding affinity as well as cleavage reaction efficiency in the left half site than the right half site (the outer region in dif appears to be determinant in the order of binding and cleavage reactions). The left half site preference is due to stronger interaction between XerH and difH left site ΔG = -21.3 kcal/mol compared to the right half site ΔG = -15.4 kcal/mol). The extra nucleotide thymine (T4) in the outer region of the left half site confers a specific hydrogen bond between the left arm and the lysine (K290) of XerH that favors stronger protein–DNA interaction with the other three outermost nucleotides, DNA bending and specific positioning of the nucleophilic tyrosine. Surprisingly, XerH assembly on difH does not induce strong DNA bending alone and it seems to require FtsK to generate the required conformational rearrangements to favor XerH DNA exchange (Bebel et al., 2016). Results obtained by Leroux et al. (2013) in difH of C. jenuni (difH_camp) demonstrated that XerH binding to either the left or right site of difH_camp resulted in similar affinities compared to the full difH_camp site possibly due to the similarity between the outer sequences of both arms. Additionally, XerH binding to difH_camp appears to be less efficient than XerS which suggests that it is less cooperative than XerS/difSL system. Additionally, these results contradict XerH binding affinities observed in H. pylori by Bebel et al. (2016) and in most tyrosine recombinases involved in chromosome resolution since it did not show any binding preference. On the other hand, unlike binding activity, asymmetrical cleavage reactions by XerH were found with a higher efficiency for bottom-strand substrates than top strand, in agreement with the results of Bebel et al. (2016) XerH recombination was also observed in vivo between two difH_camp sites located on the same plasmid; it is also suggested that XerH might be involved in decatenation processes because of the apparent absence of Topo IV proteins in H. pylori (Debowski et al., 2012; Leroux et al., 2013). Interestingly, despite difH_camp and difSL similarities in the recombination sites, the recombinases do not cross-react (XerH does not bind difSL and XerS does not bind difH_camp sites) (Leroux, unpublished).

The Archaea dif/XerA Model

In archaea, chromosome resolution appears to be catalyzed by a single recombinase (XerA) in a FtsK-independent manner that acts on a dif-like site located in the replication terminus region (Cortez et al., 2010; Serre et al., 2013). XerA shares a conserved C-terminal domain where the active tyrosine and the conserved catalytic residues (R-K-H-R-[H/W]-Y) reside. XerA proteins are well conserved between the archaeal species analyzed with 85% of sequence similarity. The xerA gene location is highly variable in archaea; some species exhibit separated xerA/dif sequences whereas some others harbor an individual xerA/dif module. Unlike most bacteria, the dif-like site is not normally located at 180° from oriC and conversely, it is located between 122° and 144° from oriC in the analyzed genomes, although the Methanosphaera stadtmanae genome showed a dif-like site at 180° from oriC. dif-like sequences consist of the traditional structure; two inverted repeat sequences of 11 bp separated by a central region of 6 bp. XerA catalyzes cleavage reactions without any detectable strand preference (Cortez et al., 2010; Serre et al., 2013).

Although archaea do not require a FtsK homolog to perform chromosome resolution, KOPS-like motifs have been found in Archaea. These KOPS-like motifs consist of four nucleotides (GTTG OR GTTC) called ASPS (ArchaeaShortPolarizedSequences) that are skewed toward dif sites, showing a similar triangle-shaped diagram observed in Bacteria of skew inversion at dif sites (Cortez et al., 2010). Serre et al. (2013) have revealed the crystal structure of XerA proteins from Pyrococcus abyssi, and Hwa Jo et al. (2016) from Thermoplasma acidophilum. Both groups reinforced the idea of cis- cleavage reaction by XerA.