Control of Directionality in Streptomyces Phage φBT1 Integrase-Mediated Site-Specific Recombination

Streptomyces phage φBT1 integrates its genome into the attB site of the host chromosome with the attP site to generate attL and attR. The φBT1 integrase belongs to the large serine recombinase subfamily which directly binds to target sites to initiate double strand breakage and exchange. A recombination directionality factor (RDF) is commonly required for switching integration to excision. Here we report the characterization of the RDF protein for φBT1 recombination. The RDF, is a phage-encoded gp3 gene product (28 KDa), which allows efficient active excision between attL and attR, and inhibits integration between attB and attP; Gp3 can also catalyze topological relaxation with the integrase of supercoiled plasmids containing a single excision site. Further study showed that Gp3 could form a dimer and interact with the integrase whether it bound to the substrate or not. The synapse formation of attL or attR alone with integrase and Gp3 showed that synapsis did not discriminate between the two sites, indicating that complementarity of central dinucleotides is the sole determinant of outcome in correct excision synapses. Furthermore, both in vitro and in vivo evidence support that the RDFs of φBT1 and φC31 were fully exchangeable, despite the low amino acid sequence identity of the two integrases.


Introduction
Bacteriophages typically insert their genomes into host chromosomes via integrase-mediated site-specific integration between two sites of attB and attP from bacteria and phage, respectively, and form attL and attR sites to establish a lysogenic state. Under inducing conditions, the phage genome is eliminated via sitespecific excision between attL and attR to convert into the lytic life cycle [1,2]. The phage encoded integrase protein is required for both integration and excision [3,4], however, this process is highly unidirectional and controlled by a recombination directionality factor (RDF or Xis) [1,3,5,6,7].
Phage-encoded site-specific integrases have been classified into two major groups: tyrosine and serine recombinases, which contain tyrosine or serine to attack DNA substrates in the active sites of the proteins [5]. The best studied member of the tyrosine recombinase group is that from E.coli phage l [8], which has been extensively investigated both biochemically and structurally for decades [8,9,10,11,12,13]. The recombination sites for l integrase are quite different, as the attB site is short (25 bp) and simple, while the attP site is relatively large (240 bp) and contains multiple binding sites for integrase, Xis, IHF (integration host factor) and Fis (factor for inversion stimulation) [14,15]. Xis is the master regulator of l recombination, and three Xis-binding sites have been found in attP [16]. DNA bending catalyzed by Xis promotes the formation of an excisive intasome, but inhibits the formation of the integrative intasome [13].
In contrast, the mechanism of serine recombinases is less well understood, and the cd resolvase is a member of the better known serine recombinases [5,17,18]. A number of phage-encoded integrases which were classified into the large serine recombinase subgroup [19], typically contain a large C-terminus which may be involved in DNA binding and directionality control [5,20]. The best studied of these integrases in vitro are those of the Streptomyces phages wC31 [21], wBT1 [22,23] as well as TG1 [24], and the mycobacterial phages Bxb1 [25] and wRv1 [4]. In each of these cases, the recombination sites of attB and attP are simple and short (less than 50 bp), and contain central dinucleotides of crossover sites which may control the polarity of the recombination [26]. To date, RDFs of phage TP901-1 [27], Bxb1 [3], wRv1 [4] as well as wC31 [1] have been identified and actively allow excision between attL and attR, and inhibit integration between attB and attP. A DNA binding assay of wC31 and Bxb1 Xis proteins strongly supported the view that Xis interacts with the att-bound integrases to change the conformation of the complexes that favor proceeding to excisive recombination [1,3,5]. Furthermore, two new reports on large serine recombinases, one of a single-molecule analysis of Bxb1 recombination revealed the molecular bearing mechanism of DNA strand exchange [28], and the second on wC31 integrase using hybrid ''phes'' recombination sites proposed a gated rotation mechanism [29]; both strongly support the ''subunit rotation'' model for exchanging DNA strands [30].
Streptomyces phage wBT1 is a temperate phage related to wC31 that integrates its genome into SCO4848 coding a putative integral membrane protein of Streptomyces coelicolor [31]. We previously established a highly efficient site-specific in vitro recombination system based on purified wBT1 integrase, and determined the minimal sites of attB and attP [23]. The biochemical mechanisms underlying synapsis, strand cleavage and rejoining were further studied, and a model in which two alternative pathways can lead to synaptic complex formation of integration was proposed [22]. Furthermore, wBT1 integrase-based methods have also been widely used in various Streptomyces strains [31,32,33,34], mammalian cells [35] and in vitro DNA assembly [36]. However, the directionality control factor of the wBT1 recombination system has not been identified to date. Here, we aim to characterize the RDF that regulates the directionality of recombination catalyzed by wBT1 integrase. The RDF protein of wBT1, which is encoded by phage gp3 gene, was sufficient to activate in vitro excisive recombination and inhibit integrative recombination.

Materials and Methods
Strains, bacteriophages and plasmids E.coli and the Streptomyces strains used in this study are described in Table S1 in File S1. Construction of phagemid wXD101 was as follows: A 3037 bp PCR product containing gp15 to gp22 of phage wC31 using primers 59-ACTCTAGAGCCGAAGGCGCCACG-CAA-39 and 59-GCGGATCCGTCGCTGGGTGGACGTAC-39 was amplified and digested with XbaI and BamHI, and inserted into the XbaI-BamHI-cleaved plasmid, pSET152, to generate a gene-targeting vector pDXM101. Plasmid pDXM101 was then introduced into Streptomyces coelicolor strain J1929 containing a wC31 lysogen by conjugation from E.coli strain ET12567/ pUZ8002. After selected on Apramycin agar, the positive colonies, mixed with both homologous recombinants targeting wC31 prophage and integrative exconjugants, were pooled. Spores of those colonies were used for the burst of phages, plated onto soft agar with indicator strain (spores of wild type Streptomyces coelicolor strain J1929) to yield plaques. The apramycin resistant phage was then obtained by transformation of the isolated phage DNA into E.coli strain DH10B and selected with antibiotics. That resulting phagemid, designated wXD101, can be maintained in E.coli as a plasmid, and conjugated into Streptomyces as an active phage. The phage genes gp23 to gp28 were deleted during the process of homologous recombination followed by phage packaging to adapt the phage genome size. Construction of plasmid-phage wXD101(X02) and wXD101(X03) by the PCR-targeting method [37] was as described in the Results and Discussion. Details of the construction of other plasmids and primers used in this study are described in Table S1 and Table S2 in File S1.

Protein expression, purification and crosslinking
Expression and purification of wBT1 and wC31 integrase were as described in our previous work [22,23]. For wBT1 excisionase (Gp3), the gp3-wBT1 gene (accession number AJ550940.2) was chemically synthesized and cloned into expression vector pET-28b (+) to generate pET-28-Bxis, that now carries a poly-His tag fused at the N-terminus of the gp3-wBT1 gene. In addition, the wC31 excisionase, the gp3-wC31 gene, was PCR amplified from phage wC31 and cloned into pET-28b (+) (see Table S1 in File S1). The procedures for protein expression, purification [23] and crosslinking [38] were performed as described previously.

In vitro recombination and relaxation assay
Standard recombination was carried out in a reaction mixture (10 ml) containing 10 mM Tris-HCl at pH 8.0, 100 mM KCl, 5% glycerol and integrase (270 nM) [23] with or without excisionase (0.25 ml),except where otherwise indicated. The final concentration of monovalent cations (K + and Na + ) was 150 mM. Reactions using linear DNA substrates were incubated at 30uC for 30 min (or as indicated) and were terminated by heat inactivation at 75uC for 10 min or treated with proteinase K at 55uC for 30 min; products were separated by electrophoresis on agarose gels (0.8%). Reactions using supercoiled plasmids for the quantification of recombination efficiency were treated with proteinase K at 55uC for 30 min followed by transformation into E.coli strain DH10B.
Topological relaxation assays were performed with supercoiled plasmids containing attB (plasmid pZLB00), attP (plasmid pZLP00), attL (plasmid pZL5816), and attR (plasmid pZL5817) (see Table S1 in File S1). The DNA substrates were incubated with wBT1 integrase and (or) excisionase for 1 hour. Reactions were carried out similar to the in vitro recombination assays; however, the reactions were terminated by heat inactivation at 75uC for 10 min and separated by electrophoresis on agarose gels (0.8%) in TAE buffer, and visualized by post-staining with ethidium bromide (EtBr). The bands between the supercoiled DNA and relaxed circles displayed a ''ladder'' of closed circular DNA species, which were likely topoisomers with a decreasing number of knots.

Plaque assay
S.coelicolor spores of lysogen J1929 harbouring wXD101, wXD101(X02) and wXD101(X03) (1610 6 cfu) were incubated in 30 ml 26YT medium at 30uC with shaking for 16 hrs. The cultures were filtered using a 0.45 mm filter membrane to obtain phage suspension, and 10 ml of the suspensions were pipetted onto Difco nutrient broth (DNB) agar with MgSO 4 (10 mM) and Ca(NO 3 ) 2 (8 mM). The soft DNB top layers (containing spores of indicator strain J1929, 1610 8 cfu) were then added to each plate, and the plates were incubated overnight at 30uC to generate plaques.

Results and Discussion
Gp3 is the RDF in phage wBT1 recombination Streptomyces phage wBT1 integrase-mediated site-specific recombination is highly efficient both in vivo [31,35] and in vitro [22,23], and has become a very useful tool in a variety of applications [32,36]. However, the recombination directionality factor (RDF) of wBT1 has not yet been identified, and limits the further development of this system. The genome organization of phage wBT1 is highly similar to that of wC31, and the major gene products are closely related [31]. Previous studies have shown that the early phage protein, Gp3, from wC31 is an RDF of recombination which activates excision and inhibits integration [1,39]. Amino acid sequence alignment of the Gp3 from wBT1 and wC31 showed that the two proteins shared 85% identity (see Figure S1 in File S1) [32]; it is likely that the Gp3 of wBT1 has the same function as Gp3-wC31 as an RDF. Thus, the gp3-wBT1 gene was chemically synthesized and cloned into the expression vector. Details of protein expression and purification are described in Materials and Methods. As shown in Figure S2 in File S1, Gp3-wBT1 was isolated with a purity over 95%, and was analyzed by both SDS-PAGE and gel filtration (data not shown).
Then in vitro excision and integration assays were performed using linearized substrates. Considering the apparent binding affinities (K d ) of integrase to the attB, attP, attL and attR sites observed at 60 nM in our previous study [22], approximately half of the concentration of integrase (27 nM) was used for primary assays ( Figure 1A and 1B). The excision products (attB and attP) could be detected when an equal amount of Gp3 (28 nM) was added, and the productivities increased gradually with increasing amounts of Gp3. An obvious excision band was observed at 140 nM of Gp3; and inhibition of integration was detected when a little Gp3 (7 nM, 1/4 of Int) was added to the reaction ( Figure 1B). This observation are consistent with the results observed in mycobacteriophage wRv1 excision, where 50% product formation was reduced by adding as little as 24 nM of RDF(concentration of integrase was 400 nM) [4]. However, it is different from that found in the wC31 recombination, that is, equal to or greater than 1:2 Gp3 to Int was sufficient to inhibit integration [1]. Our data indicated that Gp3 monomer might interact with Int tetramer in the synaptic complex to inhibit integration.
To determine the polymeric form of the Int and Gp3, crosslinking experiments were performed using purified proteins. As shown in Figure 2, both the Int and Gp3 formed dimers in solution; and when the two proteins were mixed together, a ladder of higher order oligomers was observed, indicating that a series of intermediate oligomers were formed. These results confirmed the protein-protein interactions of Int and Gp3 in solution; and indicated that these interactions could occur with different numbers of Int and Gp3 molecules, before obtaining the appropriate stoichiometry.
To investigate the kinetics of excision, the time course of in vitro excision was studied. The excision recombination showed relatively slower kinetics than that of integration [23], where 50% of products were observed in 0.5 hours; however, excision took 2 hours ( Figure 1D). Furthermore, our previous study showed that integrase could catalyze the topological relaxation of supercoiled plasmids containing single integration sites (attB or attP) in a partner DNA-independent manner, but not for the excision sites (attL or attR) [22]. Therefore, topological relaxation assays of plasmids containing attB, attP, attL or attR were performed with or without Int or Gp3( Figure 1E). Topological relaxation catalyzed by wBT1 integrase using supercoiled attL or attR plasmids was observed following the addition of Gp3; and no inhibition of relaxation on supercoiled attB or attP plasmids was observed ( Figure 1E), indicating that the presence of Gp3 had no effect on the DNA binding and cleavage activity of integrase on the single integration sites.

DNA binding properties of Gp3
To test the ability of Gp3 to bind to substrates attB, attP, attL and attR, electrophoretic mobility shift assay (EMSA) was performed using FAM-labeled attB 212 , attP 247 , attL 306 and attR 153 . As shown in Figure 3, no DNA binding activity was detected when Gp3 was the only protein in the reactions, including at a high concentration of 1750 nM. Thus, similar to the RDFs of phage Bxb1 and wC31 [1,3], the Gp3 of wBT1 had no DNA binding ability to the target substrates.
We further explored the properties of integrase bound to DNA in the presence of Gp3, since protein-protein interactions were observed in Figure 2. As shown in Figure 3A and 3B, integrase could bind to attB or attP to form ''Complex I''; and after Gp3 was added, a slower migrating complex (complex II) was detected. This observation suggested that Gp3 might interact with integrase and form a stable complex to inhibit integration between attB and attP [1,3,5]. Nevertheless, the slower migrating complex (complex II) was not observed when substrates attL or attR were tested( Figure 3C and 3D); this result was consistent with the DNA binding properties in Bxb1 excision, where the RDF of Bxb1 might bind weakly to the attL/R-Int complexes and fail to be detected during gel electrophoresis [3]. However, slower migrating complexes were detected when integrase and Gp3 of wC31 were incubated with attL or attR [1]. It is interesting that although the RDF for wBT1 shared much higher identity of the amino acids sequences with wC31 than that of mycobacteriophage RDF for Bxb1, the effects on the conformations of attL/R-Int complexes may be diverse, or scarcely show any association with similarity of the amino acids sequences.

Complementarity of central dinucleotides is the sole determinant of recombination outcomes in the correct excision synapses
Interestingly, a very slow-migrating band was formed, as shown in Figure 3C and 3D, when single substrate attL or attR alone was incubated with Int and Gp3 in the reaction. However, this was not observed in the excision of wC31 [1]. This might be a consequence of attL or attR synapses with itself. It raised a question regarding attL or attR that each could be involved in recombination with itself. Thus, excision reactions were performed using substrates containing palindromic central dinucleotides (attL GC and attR GC ). As shown in Figure 4, both attL GC and attR GC could participate in recombination with itself, and the unnatural products were generated by anti-parallel alignment of the substrates, i.e., BOB and P9OP9 for attL, POP and B9OB9 for attR ( Figure 4B). Furthermore, there is the possibility of generating recombination products by parallel alignment of substrates; however, as shown in Figure 5, the relative recombination efficiencies of attL/attL and attR/attR were only 4% and 3%, respectively. Thus, these data supported the view that the slow-migrating band shown in Figure 3C and 3D was accumulation of synaptic complexes caused by anti-parallel alignment of attL or attR with itself, which contained 59-GT central dinucleotides and further religation was suspended in the synapsis.
To gain further insights into the substrate alignments and formation of correct excision synapses in the recombination, attL GC and attR GC were incubated with integrase and Gp3 in one reaction. As shown in Figure 4, the natural excision products of attB and attP were detected; however, the unnatural products were formed with equal efficiency. Thus, the products for three forms of anti-parallel alignment of the substrates, i.e., attL/attL, attR/attR and attL/attR, were observed; and the product (BOP and B9OP9) for parallel alignment of attL and attR was detected with very low yield. Considering the low selectivity to the arm sequences of attB and attP in our previous study [22], it is believable that synapsis does not discriminate between attL and attR as represented in Bxb1 excision [40]. Thus, complementarity of central dinucleotides is the sole determinant in recombination outcomes once the correct excision synapses are formed. This property is consistent with that observed in excision recombination of Bxb1, wRv1 and wC31 [1,20,40].

Substrate selection in the presence of integrase and Gp3
To address the substrate specificity of excision recombination, we used a quantitative in vitro excision assay. Reporter plasmid substrates were constructed in which the target sites were inserted in-frame into lacZa of plasmid pBC-SK (2). Following in vitro recombination with partner plasmid, the products were transformed into E.coli and plated on indicator media: white colonies indicated that recombination had occurred. As demonstrated in Figure 5, the frequency of white colonies was more than 90% when integration occurred between attB and attP; however, the efficiency was reduced to 2% in the presence of Gp3. Typical excision catalyzed by integrase and Gp3 between attL and attR gave 89% white colonies. No substantial amounts of white colonies were observed between other combinations of target and partner plasmids, although those between attL/attL (4%), attR/attR (3%) and attB/attL (3%) showed relatively higher efficiencies. However, it seems that excision of wBT1 showed more stringent substrate selectivity than that of wC31, where white colonies of 36% for attL/attL and 18% for attR/attR were observed [1].
Cross-functional excision of integrase with Gp3 from wBT1 and wC31 respectively Since the Gp3 proteins from phage wBT1 and wC31 shared 85% identity of amino acid sequences (see Figure S1 in File S1), it is reasonable to believe that the two proteins were exchangeable during in vitro recombination. The gp3-wC31 gene was then PCR  amplified from phage wC31 and cloned into the expression vector, and the Gp3-wC31 was purified to near homogeneity (see Figure  S2 in File S1).
In vitro excision assays were performed both using linearized substrates of wBT1 and wC31. As shown in Figure 6, both Gp3-wBT1 and Gp3-wC31 could catalyze efficient excision between attL/attR-wBT1 or attL/attR-wC31 with their corresponding integrases. It is interesting that the wBT1 and wC31 integrases only showed 26% identity of amino acid sequences; their RDFs, however, shared very high identity (85%) and were proved replaceable through in vitro recombination assays ( Figure 6). This suggested that the protein structures to interact with RDFs of wBT1 and wC31 integrases might be highly similar despite the low identity of amino acid sequences. Furthermore, the Streptomyces phage TG1 [24,41] protein Gp25 (26.4 KD) shared 62% identity with Gp3-wBT1, and 60% of that with Gp3-wC31(see Figure S1 in File S1); suggesting the similar function of Gp25 in TG1 recombination. Furthermore, among four previous experimentally identified RDFs for large serine recombinases, Xis of the mycobacteriophage wRv1 (GpRv1584c, 8KD) [4] and lactococcal phage TP901-1 (Orf7, 7.5 KD) [27] are relatively small; Gp3 (27 KD) of Streptomyces phage wC31 [1] and Gp47 (28 KD) of mycobacteriophage Bxb1 [3] are larger. And they shared no identity of amino acid sequences, which is consistent with the diversity of RDFs [7].  Gp3 of wBT1 could serve as the RDF in wC31 excision out of the host genome To confirm the cross-functional excision property of integrase with Gp3 from wBT1 and wC31 in vivo, plaque assays of mutated phage wC31 were performed. Neither integration nor excision is typically essential for lytic propagation following phage infection; however, excision is required for productive lytic growth from a prophage. To test this, we constructed lysogen harboring wC31 with gp3 deletion, wondering if the phage release will decrease and that mutation could be complemented by gp3-wBT1.
A wC31 derivative phagemid wXD101 was generated, which was maintained in E.coli as a plasmid, and conjugated into Streptomyces as an active phage. Details of the construction of wXD101 are described in Materials and Methods. As illustrated in Figure 7A, the gp3-wC31 gene of wXD101 was replaced by Chloramphenicol resistant gene (Chl R ) to generate wXD101(X02) using the PCR-targeting system [37]; and further replaced by gp3-wBT1 and Chl R to obtain wXD101(X03). PCR analysis using primers X09/X12 confirmed the successful construction of the three plasmids (see Figure S3A in File S1). The plasmids were then introduced into Streptomyces coelicolor indicator strain J1929 by conjugation, and Figure S3B in File S1 shows the PCR verification of the positive exconjugants of wild-type S.co J1929 and harboring wXD101, wXD101(X02) or wXD101(X03).
As shown in Figure 7B, plaques were clearly formed when using phage suspension from strains harboring gp3-wC31 as well as gp3-wBT1; however, no plaques of the gp3-wC31 deletion construct were detected. The quantitative analysis of pfu/bacteria is shown in Figure 7A. The phenomenon that no plaques detected on the second plate was over expected. It seems like the RDF Gp3-wC31 or Gp3-wBT1, is required both for prophage excision and phage DNA replication, as described in Bxb1 RDF Gp47 [42]. This observation supported that Gp3-wBT1 shared identical function with Gp3-wC31 in vivo, could serve as the RDF in wC31 excision out of the host genome. Thus, Gp3-wBT1 and Gp3-wC31 are interchangeable in both in vitro and in vivo recombination; this could be a major concern when combining these two systems for genetic manipulation.
In conclusion, we have demonstrated that the phage-encoded protein, Gp3, is the RDF which controls the directionality of the reaction in wBT1 integrase-mediated site-specific recombination; and this function is realized by a protein-protein interaction with the integrase rather than direct binding to the substrates. Furthermore, the wBT1 integration system has been widely used Figure 5. Quantification of in vitro excision efficiencies between two sites of attB, attP, attL or attR. In vivo detection of in vitro recombination products by blue-white screen; the four recombination targets were inserted into lacZa such that LacZ activity was abolished by recombination. The four target plasmids were pZLB00 (atttB), pZP00 (atttP), pZLL00 (atttL) and pZLR00 (atttR); the plasmids containing partner DNA were pZL5812 (attB), pZLP00 (attP), pZL5816 (attL) and pZL5817 (attR). After in vitro recombination, the plasmids were transformed into E.coli DH10B and plated on IPTG/X-gal medium; the white clones were produced by recombination. The substrate ratio of each combination was 5:1 (partner to target). The data shown are average values of three reactions. The substrates used in each reaction and the relative recombination efficiencies (%) are indicated. doi:10.1371/journal.pone.0080434.g005 Figure 6. In vitro excision using Gp3 from wBT1 and wC31. Excision reactions were performed using attL and attR of wBT1 or wC31. attL and attR of wBT1 were obtained as described in Figure 1. For substrates of wC31, plasmid pZL5822 was digested with EcoRI to generate attL-wC31(1969 bp) and attR-wC31 (6270 bp); the product sizes were predicted as 3432 bp for attB wC31 and 4810 bp for attP-wC31. Pairs of excision substrates were incubated with their own integrase in the presence or absence of Gp3 from wBT1 or wC31. The concentrations of proteins used were 270 nM for integrase and 350 nM for Gp3. The reactions were terminated by proteinase K after incubation at 30uC for eight hours. doi:10.1371/journal.pone.0080434.g006 for genetic engineering both in vivo [32,33,34,35] and in vitro [22,23,36]; thus identification of the RDF (Gp3) reported here, could extend the potential utility of the wBT1 recombination system [32].