Bacteriophage-mediated horizontal gene transfer enhances bacterial adaptive responses to environmental changes, and it is one of the mechanisms responsible for the rapid spread of antibiotic resistance, bacterial virulence and pathogenicity (Canchaya et al., 2003; Brussow et al., 2004; Brown-Jaque et al., 2015; Penades et al., 2015; Touchon et al., 2017). Bacteriophages, or simply phages, play active roles in the specialized mobilization of discrete chromosomal regions (specialized transduction), and also with significant efficiency can transfer any chromosomal segment or plasmid DNA (generalized transduction). The difference between these two transduction modes is that specialized transduction is the consequence of the faulty excision of the prophage from the bacterial chromosome, resulting into packaging of phage DNA as well as adjacent DNA from the bacterial chromosome (Canchaya et al., 2003; Brussow et al., 2004; Penades et al., 2015; Touchon et al., 2017). In generalized transduction, phage DNA mispackaging occurs, and the viral packaging machinery uses chromosomal or plasmid DNA as a substrate for DNA packaging into the empty proheads instead of viral DNA (Ikeda and Tomizawa, 1965; Viret et al., 1991). Generalized transduction, which is recognized as a widespread mechanism for the transfer of any gene from one bacterium to another, was originally reported in γ-proteobacteria (Zinder and Lederberg, 1952; Lennox, 1955), and it has been also reported in many Gram-positive pathogens (Maslanova et al., 2013; Giovanetti et al., 2014; Winstel et al., 2015). The majority of generalized transducing phages package their DNA by the headful packaging mechanism (pac phages). One remarkable event related to this, is the encapsidation of pathogenicity islands, as it occurs with the Staphylococcus aureus pathogenicity islands (SaPIs). SAPIs have developed elegant strategies to hijack the phage machinery to use it for their own transfer (Penades et al., 2015). Most SaPI helper phages identified to date are pac phages, and many well-studied SaPIs are packaged by the headful mechanism (Ruzin et al., 2001). Despite its importance in spreading antibiotic resistances and virulence, the mechanisms that occur inside the cell and lead to the erroneous encapsidation of foreign DNA upon phage infection remain largely unexplored.

SPP1 is a 44-kb virulent Bacillus subtilis phage that can carry out generalized transduction (plasmid and chromosomal) with a significant frequency (Yasbin and Young, 1974; Ferrari et al., 1978; Canosi et al., 1982). The SPP1 replication and packaging machineries have been studied in deep (Alonso et al., 2006; Lo Piano et al., 2011; Oliveira et al., 2013). SPP1 DNA replication starts by the theta mode when the replisome organizer, G38P, binds to the replication origin, oriL (Pedre et al., 1994; Missich et al., 1997; Seco and Ayora, 2017). Then, the phage helicase loader (G39P) recruits the replicative haxameric helicase (G40P). The viral helicase recruits the host-encoded primase (DnaG) and DnaX, which is a subunit of the clamp loader (Pedre et al., 1994; Ayora et al., 1999; Martinez-Jimenez et al., 2002), so that a full replisome is loaded at the phage origin. SPP1 replication uses the host replicase holoenzyme and topoisomerases from the host (Seco et al., 2013; Seco and Ayora, 2017). After one or two rounds of theta-type replication (TR), it shifts to concatemeric (sigma-type) DNA replication in a process driven by recombination (Lo Piano et al., 2011). Two viral proteins may participate in this shift, the ATP-independent single-strand annealing recombinase (G35P) and its partner, the 5′→3′ exonuclease (G34.1P) (Ayora et al., 2002; Martinez-Jimenez et al., 2005). In the shift to concatemeric DNA replication, G38P, bound to oriR, or working as a pre-primosome organizer (like the bacterial PriA enzyme), may restart DNA replication at stalled or paused replication forks (Seco et al., 2013). SPP1 codes for two other proteins involved in DNA replication and recombination: the G36P and G44P proteins. G36P is a single-stranded DNA binding protein (SSB), and G44P is a Holliday junction resolvase of the RusA family, which recognizes and cleaves a variety of recombination intermediates (Martinez-Jimenez et al., 2005; Zecchi et al., 2012). Biochemical assays showed that G36P is crucial for SPP1 DNA replication in vitro, but it can be substituted by host-encoded SSB (known as SsbA) (Seco et al., 2013). The role of G44P in SPP1 replication is thought to be the processing of the stalled replication fork, which may trigger the shift to the sigma-type or concatemeric DNA replication. This type of DNA replication is essential to generate the concatemeric DNA, which is the substrate for encapsidation. Viral replication and packaging are sequential and in some way coupled events. SPP1 encapsidates linear double-strand (ds) DNA into an empty prohead by a processive (∼4 sequential packaging cycles) headful packaging mechanism, using the linear head-to-tail concatemer as a substrate (Oliveira et al., 2013). This is consistent with the observation that an in vitro DNA packaging system efficiently packaged mature SPP1 DNA as well as linear plasmid DNA, but no DNA packaging could be detected when circular DNA was the substrate for encapsidation (Oliveira et al., 2005). SPP1 packaging is initiated with the recognition of the specific pac region by the terminase small subunit, G1P, and the sequence specific cleavage at the pac sequence (CTATTGCGG↓C) by the terminase large subunit, G2P (Chai et al., 1992, 1995, 1997). This generates the first DNA end to be encapsidated (Chai et al., 1992; Gual et al., 2000; Camacho et al., 2003). A sequence independent cleavage, at 104% of the genome (headful cleavage), terminates one packaging round, generating a new starting point for another one (Chai et al., 1995; Camacho et al., 2003). Hence, the first cleavage in the concatemeric SPP1 DNA occurs specifically at pac, whereas the next ones do not (Gual et al., 2000).

In addition to package viral DNA, SPP1 is able to encapsidate chromosomal or plasmid DNA. However, some differences were observed with these two substrates. Rolling-circle replicating plasmids could be transduced at a frequency much higher than chromosomal DNA (Ferrari et al., 1978; Deichelbohrer et al., 1985), and an explanation for this could be that the copy number of plasmids in the cell is higher than that of the chromosome. Alternatively, another possibility could be that the replication mode influences the transduction frequency. It was also observed that the frequency of transduction of pUB110 and pC194 naturally occurring plasmids was enhanced 100- to 1000-fold by the presence of inserts homologous to the transducing phage DNA (Deichelbohrer et al., 1985). This homology-facilitated plasmid transduction was independent of the host RecA (Canosi et al., 1982; Deichelbohrer et al., 1985). In contrast, another report showed that SPP1 mediated chromosomal transduction was reduced 30-fold in cells having mutations in host functions involved in homologous recombination, such as RecA, RecU, and RecF (Ferrari et al., 1978). These differences, which were observed between plasmid and chromosomal transduction in the SPP1 system motivated us to analyze in deep and throughout the manuscript the influence of the replication mode and of the plasmid copy number in plasmid generalized transduction. In addition, we have analyzed the role of phage recombination and replication proteins. We show that in absence of G35P, G38P, G39P, or G40P linear plasmid transduction is blocked. In contrast, establishment of injected linear plasmid DNA in the recipient host was independent of viral-encoded functions. The transfer efficiency was found to be dependent on homology to phage DNA, plasmid copy number, and replication mechanism.

Until recently, it was thought that generalized transduction occurred at low frequency. However, recent single-cell analyses observed transduction rates close to 1% per plaque forming units when natural communities were used as recipients (Kenzaka et al., 2010). Therefore the study of the transduction mechanisms is essential to prevent this highly frequent horizontal gene transfer process, to avoid the spread of antibiotic resistance among bacteria. In this aspect, the SPP1 bacteriophage is a valuable model, because its replication, recombination, and packaging machineries haven been studied in deep for many years. Furthermore, it was recently reported that SPP1 can occasionally infect resistant cells when combined with sensitive cells, providing new routes for horizontal gene transfer (Tzipilevich et al., 2017). Previous biochemical studies assigned a role to SPP1 proteins G34.1P, G35P, G36P, G38P, G39P, G40P, and G44P in replication and recombination, but their contribution to generalized plasmid transduction remained unknown. Here we show that all SPP1 replication proteins contribute to horizontal plasmid transfer, although to a different extent. The origin binding protein (G38P), helicase loader (G39P), and helicase (G40P) are essential to produce concatemeric plasmid DNA, which is synthesized after phage infection. Infections with the sus36 mutants show only a 10-fold reduction in the transduction frequency, probably due to potential complementation of the G36P function by cellular SsbA protein (Seco et al., 2013; Seco and Ayora, 2017). The SPP1 recombination proteins contribute to plasmid transfer to a different extent. The exonuclease G34.1P and the Holliday junction resolvase G44P only contribute partially to plasmid transduction, with a reduction of the transduction frequency of 12- and 5-fold in their mutants, respectively. The G35P recombinase is essential, with its inactivation leading to a >100-fold decrease.

Previous studies with SPP1 and RCR plasmids showed that: (i) the transduction of pUB110 and pC194 plasmids was enhanced 100- to 1000-fold when there was any homology between the plasmid and the SPP1 genome rather than with the specific pac signal; (ii) pUB110 and pC194 plasmid transduction was independent on RecA (Deichelbohrer et al., 1985), and (iii) linear plasmid concatemeric DNA (or high-molecular-weight [hmw] DNA) accumulated during phage infection, and in certain genetic backgrounds (Viret and Alonso, 1987; Viret et al., 1991). The synthesis of hmw DNA and its independence of the host-encoded recombinase (RecA) strongly suggests that the formation of transducing particles may rely on viral replication and/or recombination functions. In this work we show that the synthesis of this hmw DNA, and consequently transduction of RCR plasmids requires an active G35P protein. Biochemical analysis shows that G35P is an ATP-independent single-strand annealing enzyme, similar to the RecT enzyme encoded by the Rac prophage (Ayora et al., 2002). Both, G35P and RecT, belong to the Redβ family of viral single strand annealing proteins. To date, five different single strand annealing recombinase families have been identified in phages: Sak, Redβ, Erf, Sak4 and Gp2.5 (Lopes et al., 2010). These recombinases have gained increased attention in recent years because of their abundance in phage genomes (Lopes et al., 2010; Delattre et al., 2016), and also due to their wide use in recombineering systems (Datta et al., 2008; Sun et al., 2015). Many of these recombinases, including G35P, are essential for the phage life cycle (Zecchi et al., 2012; Neamah et al., 2017).

In this work we found that variations in copy-number affect the transduction frequency. Since the transduction is a stochastic process, it is expected that the more plasmid DNA in the cell the more generalized transducing phage particles should carry a plasmid copy and therefore the chances of transduction increase. The plasmids replicating in B. subtilis cells are either of the TR (circle-to-circle) type or RCR (sigma) type, and the products of both replication modes are usually covalently closed circular monomers (Khan, 2005). Comparing plasmids with similar copy number we observed that the frequency of transduction for RCR plasmids is ∼60-fold higher than that for TR plasmids. This result suggests that the type of DNA replication also determines the transduction frequency. In the small RCR plasmids leading and lagging strand replication are uncoupled, and they contain two modules: the Rep protein with its cognate double-strand origin (DSO), and a single strand origin (SSO), which functions as the major initiation site for lagging-strand synthesis (Alonso et al., 1988; Espinosa et al., 1995; Khan, 2005). All RCR plasmids accumulate ssDNA although to a different extent: pUB110 accumulates traces and pC194 accumulates circular ssDNA (te Riele et al., 1986; Viret and Alonso, 1988). In contrast, the large low-copy-number TR plasmids, such as pBT233, which replicates via an unidirectional mechanism, do not accumulate circular ssDNA intermediates (Ceglowski et al., 1993b,c). We propose that the high transfer frequencies of some RCR plasmids may be correlated with the high accumulation of recombinogenic ssDNA intermediates in these plasmids. Such ssDNA intermediates may constitute the substrates for formation of the transducing particles, through a recombination catalyzed by the G35P protein. This is in agreement with recent results observed with viral recombinases: when analyzing their recombineering activity in vivo, it was found that they catalyze single-strand annealing preferentially on the lagging strand (van Kessel and Hatfull, 2008; Mosberg et al., 2010; Lajoie et al., 2012; Fricker and Peters, 2014; Ander et al., 2015). We propose that all the phages encoding recombinases will transduce RCR plasmids with high efficiency by the mechanism of viral recombinase-mediated generalized transduction. Furthermore, we also observed that the transduction of the pUB110 and pNDH33 plasmids, which do not have an extensive region of homology, was strongly reduced in infections with the sus35 mutant (Figures 2, 4). All phage recombinases studied so far are single-strand annealing proteins that promote genetic recombination under more permissive conditions than RecA (Scaltriti et al., 2011; De Paepe et al., 2014; Menouni et al., 2015). Our results suggest that G35P contributes to the transfer of natural plasmids by catalyzing a recombination reaction using small stretches of homology found in many plasmids (Supplementary Table S1).

The different contributions of the SPP1 recombination proteins to plasmid transduction, together with the high recombinogenic nature of the RCR plasmids, suggest that the initial DNA substrate, used for the production of transducing particles by recombination, is indeed ssDNA. This is consistent with the result that the G34.1P exonuclease, which resects the dsDNA ends to generate the appropriate substrate for the recombinase (Martinez-Jimenez et al., 2005), has a minor role in plasmid transfer. Similarly, we found that the SPP1 SSB protein, G36P, only slightly contributes to the mechanisms of plasmid transduction. However, in some phages the recombinases require the activity of their cognate SSB proteins to perform their function (Neamah et al., 2017).

It was previously observed with RCR plasmids that any SPP1 DNA segment larger than 50 bp, cloned into such plasmids, greatly increased the transduction frequency (Deichelbohrer et al., 1985; Alonso et al., 1986). We extend this observation to TR plasmids, where the transduction frequency was highly increased, independently of what is the region of homology cloned, whether it was the packaging sequence, a phage origin of replication, or any other region of homology. Similarly, the cloning of the origin of replication of SPP1 (oriL) into a RCR-type plasmid did not further increase its transduction frequency (Supplementary Table S2). We conclude that any DNA region homologous to the phage genome increases the frequency of horizontal transfer of plasmids, independently of their replication mechanism. Enhanced transduction of plasmids bearing homology with phage DNA has been also observed with phage T4, which codes for a different recombinase, the UvsX protein (Kreuzer et al., 1988), and with Salmonella typhimurium phage P22, which codes for the Erf recombinase (Orbach and Jackson, 1982).

How is the plasmid substrate for generalized transduction generated? Three different mechanisms could account for the generation of a concatemeric plasmid DNA with high frequency of transduction. In the first model, the multiple tandem repeats of plasmid DNA might be produced by intermolecular recombination, as proposed for P22 plasmid transduction (Mann and Slauch, 1997). This mechanism resembles phage T4 generation of concatemeric DNA during its replication (Kreuzer, 2000; Mosig et al., 2001). Here, multiple strand invasions catalyzed by the ATP-dependent RecA-like recombinase encoded by this phage, UvsX, and the resolution of the Holliday junction intermediates by its Holliday junction resolvase Gp49 (also called EndoVII), produce the concatemeric DNA, as well as the transducing particle (Kreuzer et al., 1988; Kreuzer, 2000; Mosig et al., 2001). We do not favor this hypothesis in the SPP1 system, because we found that the Holliday junction resolvase G44P has only a minor role in plasmid pBG55 and pUB110 transduction. In the second model, plasmid over-replication leads to the accumulation of linear concatemeric hmw DNA (Cohen and Clark, 1986; Viret and Alonso, 1987; Viret et al., 1991). The accumulation of linear head-to-tail multigenome-length plasmid DNA (hmw DNA) in the absence of RecBCD/AddAB was documented in both Escherichia coli and B. subtilis cells (Silberstein and Cohen, 1987; Viret and Alonso, 1987). Indeed, upon infection, many bacteriophages directly or indirectly inactivate end-resection catalyzed by this host encoded multi-subunit helicase-nuclease enzyme (Szczepanska, 2009). It was observed that the synthesis of pC194 or pUB110 hmw plasmid DNA occurred in the absence of plasmid-encoded Rep protein, and required DNA PolI, RecA and pre-primosomal proteins (e.g., DnaB) (Viret and Alonso, 1987; Leonhardt et al., 1991; Viret et al., 1991). Analysis of this hmw plasmid DNA by electron microscopy displayed linear DNA molecules up to 100 kb in size, which were either single-stranded, double-stranded or duplex DNA with single-stranded tailed ends (Leonhardt et al., 1991). This hmw DNA can be encapsidated into a viral prohead by a headful packaging mechanism (Schmidt and Schmieger, 1984; Schmieger, 1984). If this model is correct, the presence of a pac signal will significantly increase the encapsidation of the plasmid hmw DNA, and we found that there was not an increase in the transduction frequency when plasmids contained the pac signal. In the third model, phage infection arrests host and plasmid replication. Then SPP1-dependent replication restarts, and the linear plasmid concatemer is synthesized. This is consistent with the result that the phage G38P protein may act as a PriA-like enzyme, restarting DNA replication outside form a replication origin (Seco et al., 2013; Seco and Ayora, 2017). In this de novo synthesis of plasmid DNA, a viral pac site might be gained by recombination and recognized by the viral packaging machinery (Alonso et al., 1986; Bravo et al., 1990; Viret et al., 1991). In this model, the phage might form a phage-plasmid chimera and the plasmid hijacks the viral replication machinery to promote de novo synthesis of linear plasmid concatemeric DNA. The concatemeric plasmid DNA is then packaged into an empty prohead by the headful mechanism, indistinguishable of viral DNA, provided that the packaged substrate is larger than mature phage DNA. Our data support the third model, because we found that in infections with a phage bearing a mutation in the terminase (sus19 infections), plasmid concatemers up to 200-kb long are produced after phage infection (Figure 3B and Supplementary Figure S2). This model explains also the requirement of viral replication proteins for the formation of the transducing particles. However, we were unable to detect the phage-plasmid chimeras, which might be rapidly processed to produce the plasmid head-to-tail concatemers.

Our results show that the establishment of the transduced concatemeric plasmid in the host is independent of phage encoded recombination functions, which only participate in the generation of the transducing particle. We propose that the injected linear concatemer can be converted into a circular form by the homologous recombination machinery of the recipient cells. In this respect, transduction of plasmids might have similar host requirements as the resolution of phage-plasmid chimeras analyzed in the P22 and SPP1 systems (Orbach and Jackson, 1982; Alonso et al., 1992). In the former case, the plasmid integrated into the phage genome has to be excised from the genome of the defective phage prior to establishment, whereas in the latter case the head-to-tail plasmid concatemer has to recombine intramolecularly to facilitate plasmid establishment. This process was found to be RecA-independent but dependent on host RecO and RecR functions that also catalyze single-strand annealing (Alonso et al., 1992; Manfredi et al., 2008).

AV-R, ML-S, AQ-O, and SA: performed the experiments; AV-R, AS, and SA: analyzed data; SA: conceived the project, integrated the results and wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.