PLE encodes a gene product (Int) with an N-terminal serine recombinase domain and a large C-terminus containing a putative zinc ribbon domain and coiled–coiled motif characteristic of large serine recombinases (LSRs)21. Typically found in temperate phages, LSRs have the ability to catalyze recombination between attachment (att) sites22,23,24. Only the LSR is required to catalyze recombination between episomal (attP) and chromosomal (attC) sites, leading to integration of the episome into the host chromosome. To reverse this process, an RDF is required to physically interact with the LSR and direct the LSR to recombine the left (attL) and right (attR) attachment sites, resulting in excision of DNA between these sites (Fig. 1b). In addition to what has been documented regarding PLE circularization following ICP1 infection under laboratory conditions18,20, we note that ICP1-dependent PLE circularization can be detected in cholera patient stool (Fig. 1c), underscoring that V. cholerae PLE responds to ICP1 infection during disease in humans.
To determine if Int plays a role in PLE circularization during ICP1 infection, PLE 1 Δint was challenged with ICP1. Unlike wild-type PLE, circularization was not detected in the PLE 1 Δint strain. PLE circularization was restored with in trans Int expression, but only during ICP1 infection (Fig. 1d, Supplementary Fig. 1a). As Int is necessary for PLE circularization to occur, we next investigated the expression pattern of Int during ICP1 infection. We introduced a FLAG-tag into the endogenous copy of int and confirmed that FLAG-tagged Int retained the ability to catalyze circularization within 5 min of ICP1 infection (Supplementary Fig. 1b). We detected FLAG-Int in uninfected cells and observed that the level of Int did not increase during ICP1 infection (Fig. 1e), showing that ICP1 infection does not induce int expression. Interestingly, although PLE 1 Δint cannot functionally circularize following phage infection, PLE 1 Δint still inhibits ICP1 plaque formation (Fig. 1f), suggesting that excision is induced separately from other components of PLE that are needed for anti-phage activity.
To determine if PLE 1 Int is a functional LSR, we performed integration assays to probe the ability of Int to recombine the chromosomal attC and PLE attP sites. Through both in vivo (Supplementary Fig. 1c) and in vitro (Supplementary Fig. 1d) assays, we found that Int was sufficient to recombine attP and attC sites, as is characteristic of LSRs22,23. These assays demonstrate that Int is necessary for PLE circularization in response to ICP1 infection and that Int is a functional LSR that can catalyze recombination between att sites.
As Int is constitutively expressed and not sufficient to catalyze the circularization of PLE 1 in the absence of phage infection (Fig. 1e, Supplementary Fig. 1a), we hypothesized that an RDF is required to direct Int to recombine the attL and attR sites as is characteristic of LSRs25,26. There are no conserved sequence characteristics of RDFs that enable homology-based identification27, however, in characterized LSR/RDF systems of temperate phages, both the LSR and RDF are encoded within the same genome25,28. To evaluate if the RDF is PLE-encoded, PLE 1 strains harboring gene cluster deletions of predicted open reading frames (ORFs) were screened for circularization defects during ICP1 infection. Unexpectedly, all of the PLE ORF knockouts still circularized, implying that the RDF is not PLE-encoded (Fig. 2a). To establish the minimal PLE-encoded factors required for circularization, we constructed a ‘miniPLE’, which has Int under control of its endogenous promoter and a kanamycin cassette, flanked by att sites, and integrated into the V. cholerae chromosome in the same location as PLE 1 (Fig. 2b). In support of the mutational analyses (Fig. 2a), the miniPLE circularized and excised from the chromosome during ICP1 infection (Fig. 2b., Supplementary Fig. 2). Together with the inability of PLE 1 Δint to circularize (Fig. 1d), these results demonstrate that Int is the only PLE gene that is necessary for PLE 1 circularization during ICP1 infection.
Due to the specificity of circularization during ICP1 infection, we hypothesized that ICP1 encodes a gene product that directs Int-mediated PLE circularization during infection. To identify this phage-encoded gene product, we screened for ICP1 mutations that abolished miniPLE circularization during infection. Through this screen, we identified a mutant phage that failed to circularize the miniPLE (Supplementary Fig. 3a). Sequencing revealed that this mutant phage had a deletion of orf50 and orf51 caused by recombination between orf49 and orf52 (sites 18761–19800 in ICP1_2004_A, Sequence ID HQ641354 [https://www.ncbi.nlm.nih.gov/nuccore/HQ641354]), leading to a unique fusion gene with a novel stop codon. We evaluated the phage-encoded gene products within this region to determine if one or more was responsible for directing PLE excision during ICP1 infection. Ectopic expression of each gene product revealed that only the hypothetical ICP1 gene product annotated as orf51 (YP_004250992 [https://www.ncbi.nlm.nih.gov/gene/?term=YP_004250992]) was sufficient to induce PLE circularization in the absence of phage infection (Fig. 3a, Supplementary Fig. 3b). We constructed a clean deletion of orf51 and found that it was indeed necessary for miniPLE circularization (Fig. 3b, Supplementary Fig 3c), though interestingly, orf51 is dispensable for ICP1 plaque formation (Supplementary Fig. 3d). As this gene product is both necessary and sufficient for Int-mediated PLE excision, we named it phage-encoded excisionase (PexA). PexA is a small protein unique to ICP1 isolates that has no sequence similarity to known proteins. Consistent with the rapid kinetics of PLE 1 circularization20, we found that PexA is expressed de novo within 5 min of ICP1 infection (Fig. 3c), leading us to hypothesize that PexA is hijacked by PLE 1 to function as the RDF for Int-mediated PLE excision.
In order to direct recombination, characterized RDFs physically interact with their cognate LSR. The ability of PexA to physically interact with Int was probed with a bacterial adenylate cyclase two-hybrid (BACTH) assay, in which LacZ expression was detected when both Int and PexA were fused to adenylate cyclase subunits (Fig. 4a). This interaction was further validated using an in vitro pulldown assay, in which PexA coeluted with 6xHis-tagged Int (Fig. 4b), showing that PexA can bind to Int in vivo and in vitro.
Recombination in characterized LSR/RDF systems requires solely the LSR, RDF, and DNA substrates24,25,28. To determine if PexA directs Int to excise and circularize PLE, in vitro excision assays were performed using PCR fragments containing the attR and attL sites and purified PexA and Int (Fig. 4c). Addition of neither PexA nor Int alone led to recombination between attR and attL; however, when Int and PexA were both added, recombination products were detected (Fig. 4c). RDFs have also been shown to block LSR-mediated integration25. Consistent with this model, we observed that addition of PexA to an attC and attP integration reaction blocks Int-mediated recombination in vitro (Fig. 4d). These data demonstrate that phage-encoded PexA is the RDF for PLE 1 Int and provide the first example of an LSR/RDF pair being encoded in different genomes.
Analysis of ICP1 genomes from a 12-year period shows that PexA is maintained and that it is 99% identical in all ICP1 isolates (Supplementary Fig. 4a). The conservation of PexA leads us to speculate that although PexA is not essential for plaque formation under lab conditions (Supplementary Fig. 3d), it is likely integral to the ICP1 lifecycle in nature, and hence it may be hijacked as the RDF to signal ICP1 infection in all PLEs. Further, all five PLEs contain a putative LSR and respond to ICP1 infection by circularizing following infection20, therefore we examined the conservation of the Int/PexA interaction. Upon ectopic expression of PexA, all PLEs, except PLE 2, demonstrated functional circularization (Fig. 5a). PLE 2 has the most diverse Int, sharing 25.7% amino acid identity with PLE 1 Int across 63% of the protein, while Int from PLE 3, PLE 4, and PLE 5 are more similar to PLE 1 Int (Supplementary Fig. 4b). Consistent with the divergence of PLE 2 Int, PLE 2 also integrates into a unique site in the V. cholerae small chromosome20 (Supplementary Fig. 4c), indicating that PLE 2 Int recognizes different att sites from the other PLEs. Altogether, this data indicates that PLE 2 evolved to recognize a unique RDF, possibly altering att site specificity in the process.
Although PexA appears to be stably maintained in the natural ICP1 isolates in our collection, the observation that PLE 2 Int has evolved to use a unique RDF indicates that PexA may not have always been a reliable cue enabling PLE to respond to ICP1 infection throughout their co-evolutionary history. We also found that PLE blocks ICP1 infection independent of Int and subsequent circularization (Fig. 1f), suggesting that PLEs may have evolved to use multiple phage products to induce activity. This model of PLE induction is in stark contrast to characterized PICI systems in which a single phage product is sufficient to activate the entire PICI excision–replication–packaging program8,9. To test if PexA plays an additional role in PLE activation, we analyzed PLE copy number for PexA-linked effects. During infection with ICP1 ΔpexA, PLE still replicated to high copy, although we did observe a mild but not significant defect relative to infection with wild-type phage (Fig. 5b) presumably because PLE is unable to excise from the chromosome and replicate quite as efficiently. Additionally, ectopic expression of pexA was not sufficient to drive PLE replication in the absence of ICP1 infection (Fig. 5b), thus PexA functions as the RDF for PLE-encoded Int to stimulate PLE excision but PexA does not appear to play a role in inducing other aspects of PLE activity. Consistent with PexA serving solely as the RDF and our previous observation that PLE Δint still blocks ICP1 plaque formation (Fig. 1f), ICP1 ΔpexA is still blocked by PLE 1 (Fig. 5c), confirming that PLE circularization is not required for PLE-mediated anti-phage activity.
