Screening for Highly Transduced Genes in Staphylococcus aureus Revealed Both Lateral and Specialized Transduction

ABSTRACT Bacteriophage-mediated transduction of bacterial DNA is a major route of horizontal gene transfer in the human pathogen, Staphylococcus aureus. Transduction involves the packaging of bacterial DNA by viruses and enables the transmission of virulence and resistance genes between cells. To learn more about transduction in S. aureus, we searched a transposon mutant library for genes and mutations that enhanced transfer mediated by the temperate phage, ϕ11. Using a novel screening strategy, we performed multiple rounds of transduction of transposon mutant pools selecting for an antibiotic resistance marker within the transposon element. When determining the locations of transferred mutations, we found that the screen had selected for just 1 or 2 transposon mutant(s) within each pool of 96 mutants. Subsequent analysis showed that the position of the transposon, rather than the inactivation of bacterial genes, was responsible for the phenotype. Interestingly, from multiple rounds, we identified a pattern of transduction that encompassed mobile genetic elements as well as chromosomal regions both upstream and downstream of the phage integration site. The latter was confirmed by DNA sequencing of purified phage lysates. Importantly, transduction frequencies were lower for phage lysates obtained by phage infection rather than induction. Our results confirmed previous reports of lateral transduction of bacterial DNA downstream of the integrated phage but also indicated a novel form of specialized transduction of DNA upstream of the phage. These findings illustrated the complexity of transduction processes and increased our understanding of the mechanisms by which phages transfer bacterial DNA. IMPORTANCE Horizontal transfer of DNA between bacterial cells contributes to the spread of virulence and antibiotic resistance genes in human pathogens. For Staphylococcus aureus, bacterial viruses play a major role in facilitating the horizontal transfer. These viruses, termed bacteriophages, can transfer bacterial DNA between cells by a process known as transduction, which despite its importance is only poorly characterized. Here, we employed a transposon mutant library to investigate transduction in S. aureus. We showed that the genomic location of bacterial DNA relative to where bacteriophages integrated into that bacterial genome affected how frequently that DNA was transduced. Based on serial transduction of transposon mutant pools and direct sequencing of bacterial DNA in bacteriophage particles, we demonstrated both lateral and specialized transduction. The use of mutant libraries to investigate the genomic patterns of bacterial DNA transferred between cells could help us understand how horizontal transfer influences virulence and resistance development.

staphylococcal genome is highly plastic, largely due to horizontal gene transfer mediated by mobile genetic elements (MGEs), such as staphylococcal pathogenicity islands (SaPIs), plasmids, and phages (3)(4)(5). Phage-mediated transduction appears to be an important route of gene transfer (4,6). Transduction is a process whereby some phages can package bacterial DNA at a low frequency and transfer this DNA between cells (6,7). Transducing phages are often temperate because they can both be integrated into the bacterial genome as prophages (lysogeny) or be lytic. Transduction can occur through several different mechanisms namely, generalized, specialized, and lateral transduction. In S. aureus, generalized transduction has been used for many years as a genetic tool and recently, the mechanism for lateral transduction was established (8)(9)(10). In contrast, we know little about specialized transduction in this organism.
Phage f 11 is one of the best-characterized staphylococcal phages and is regularly used as a laboratory tool for transferring genes by transduction (11). Phage f 11 is a pac-type phage, meaning that it recognizes a phage 'pac' site and packages DNA into the capsid in a 'headful' manner, terminating when the capsid is filled with slightly more than the ;45 kb phage genome (11,12). As such, f 11 can transfer bacterial DNA by generalized transduction when bacterial DNA contains sequences with homology to the phage 'pac' sites (11,13,14).
Lateral transduction is a recently identified form of phage-mediated transduction in both S. aureus and Salmonella (10,15). Here, it was found that for some temperate phages, replication begins before excision from the host chromosome leading to the replication and packing of the integrated prophage together with the flanking regions of bacterial DNA. In practice, this has led to transduction of up to 100 kb of bacterial DNA downstream of the prophage with frequencies of 1000-fold higher compared to generalized transduction. The study identified that this effect was unidirectional from the pac site and so the region upstream of the phage was not transferred (10).
Specialized transduction occurs when a prophage excises incorrectly from the bacterial chromosome (16)(17)(18)(19)(20)(21)(22)(23)(24)(25). Instead of a precise excision of the full phage genome between the att L and att R sites, an incomplete portion of the phage excises alongside some of the flanking bacterial DNA either upstream or downstream of the phage. If the phage pac site is present, this hybrid DNA will be replicated, packaged into phage capsids, and transduced (16,18,21). It is limited to the immediate regions of bacterial DNA flanking a prophage. Specialized transduction mechanisms have been reported for phages infecting several bacterial genera, including Bacillus, Salmonella, Pseudomonas, and Vibrio (16,17,20,23). However, literature is absent on specialized transduction by staphylococcal phages.
In a recent publication, DNA short-and long-read sequencing was used to investigate the patterns of bacterial DNA in transducing particles in different genera and species of bacteria (26). The patterns observed indicated that different phages and prophages are either more or less prone to lateral, specialized, or generalized transduction and that the same temperate phage may transduce bacterial DNA differently depending on whether that phage is undergoing lytic infection or lysogenic induction (27).
To understand more about transduction processes in S. aureus, we explored a screening methodology that allowed us to identify S. aureus transposon mutants that were preferentially transduced from a pool of mutants. We utilized the Nebraska transposon mutant library (NTML) in S. aureus strain JE2 (28,29), infecting pools of 96 mutants with the transducing, lysogenic phage f 11. From this screen, we anticipated identifying genes that either, when inactivated, promoted transduction or were located within the chromosome in regions transduced at higher rates than elsewhere. We showed that the transposon mutations in the chromosomal regions flanking the phage integration site were transduced at higher rates by both lateral and specialized transduction. This study established a simple screening method that reveals patterns of transduction at the genome level, which can be readily applied to transposon mutant libraries of other bacteria. Remarkably, this study was the first to demonstrate specialized transduction in S. aureus.

RESULTS
Transduction selection preferentially identifies transposon mutations in MGEs and regions flanking prophage /11. Intending to identify genes or mutations that increase transduction frequency, we screened the S. aureus JE2 NTML for mutations that were preferentially transduced. To this end we carried out consecutive rounds of transduction on pools of 96 library mutants, infecting these with the transducing temperate phage f 11 to produce an 'infection' phage lysate containing transducing particles ( Fig. 1). We then performed transduction using these lysates with the laboratory strain 8325-4 f 11, a strain containing only the prophage w 11, as the recipient. Having the f 11 prophage present in the recipient, rather than using the phage-less 8325-4, prevented phage superinfection and, thus, increased survival of the transductants. Likewise, the use of this recipient prevented a mixed population of 8325-4 transductants and 8325-4 w 11 transductants, which would occur using an 8325-4 recipient due to the integration of w 11. The transfer from JE2 to the 8325-4 w 11 was conducted to exclude interference from the prophages USA300w Sa2 and USA300w Sa3, present in JE2. The core genomes of JE2 and 8325-4 have an identity of 96.6% when aligned using MAUVE, and there was no difference in w 11 phage titer between 8325-4 and JE2 recipients (Fig. S1).
Transposon-containing transductants in 8325-4 w 11 were selected for erythromycin resistance, encoded by the transposon. These transductants were pooled and the f 11 prophage was induced to produce a 'round 1 induction' lysate. The transduction and induction steps were repeated to produce 'round 2' and 'round 39 transduction lysates. Although this was necessary for the selection of highly transduced mutations, this could also have impacted our screen by selecting for transposon mutants of greater fitness. At each stage, the colony forming unit (CFU)/mL and plaque forming unit (PFU)/ mL were recorded, and the transduction frequency was calculated, controlling for phage variation ([CFU/mL]/[PFU/mL]; Fig. S2). Importantly, the transduction frequencies showed a significant increase over the three rounds of transduction screening for the seven 96-well plates examined, with the overall mean increasing from 26.8 at infection to 25.7, 24.5, and 23.7 at rounds 1, 2, and 3, respectively (Fig. S2).
At the round 3 transduction, we collected 10 colonies from each plate pool and identified the locations of the transposon mutations as described (28). Most pools contained one predominant transposon mutant per biological replicate, leading to #4 being identified in total for each plate (Fig. 2). For the first biological replicate, only one transposon mutant was identified in 6 of the 7 pools while transposon mutants were identified for plates 1 and 2. Strikingly, half (4 out of 8) of the transposon mutations identified in our initial screen were situated on MGEs (Fig. 2). Of these, one transposon was in a SaPI5 gene, one in the USA300f Sa2 prophage of JE2 while the other 2 were located in the USA300f Sa3 JE2 prophage. These MGE-located transposon prophage to produce the next round's lysate. At round three, the resulting transductants were processed as previously described to identify the location of transposon insertion (28,34). Created with BioRender.com. mutations were distributed throughout the chromosome (Fig. 2). The other 4 chromosomally situated non-MGE-related transposon mutations clustered around the integration site of the f 11 (att B ; Fig. 2). As a prophage, w 11 integrates into an intergenic region of the 8325 chromosomes with a specific directionality, where the attL was situated at ;1.967 Mb and the attR at ;1.923 Mb (29). Relative to the f 11 prophage, 2 of the transposon mutations were situated upstream, but close to, the f 11 att B , while the other 2 were located downstream of this site. For subsequent biological replicates, the final transduction plate colonies were pooled for DNA extraction and transposon location identification, in most cases giving one predominant transposon mutant (Fig. 2). The transposon mutations identified were in some cases identical to the first replicate, while all others were in the same regions of either MGEs or upstream and downstream of the f 11 att B . To check the impact of fitness on our screen, we performed growth curves on two of the transposon mutants identified (Tn_1868 and Tn_1849) together with two mutants that were included in the pool but not found in the screen (Tn_1799 and Tn_1883) as well as 8325-4 and 8325-4 f 11 controls (Fig. S3). There were no significant differences in doubling the time between the screen-identified mutants and the nonidentified ones (Table S3). Therefore, we hypothesized that we were identifying transposon mutations that were preferentially transduced due to either induced MGE mobility or lateral and specialized transduction.
Transposon mutants upstream and downstream of /11 have an increased transduction frequency. The chromosomal transposon mutants clustering around the f 11 att B interested us as potential incidents of specialized or lateral transduction (replicate 1: SAUSA300_1866, SAUSA300_1868, SAUSA300_1849, and SAUSA300_1797). Therefore, we transduced these individual transposon mutations into 8325-4 f 11 and confirmed by multiplex PCR targeting the phage integrases that none of the JE2 endogenous phages had been transferred (30). Using these 'clean background' strains, we confirmed that the transduction frequencies in the 4 chromosomal transposon mutants were higher than our 8325-4 f 11 rotB::erm transposon control. The rotB transposon mutant has a wild-type phenotype in respect to bacterial growth and transduction, and f 11 should only transfer this transposon mutation by generalized transduction based on its position in the 8325 genomes (1.794 Mb; Fig. 2). All transposon mutants had significantly higher transduction frequencies than the rotB control (Fig. 3). The initial data (CFU/mL and PFU/mL) for these experiments can be seen in Fig. S4 and a schematic showing the locations of the transposon mutants used for Fig. 3 to 7 can be observed in Fig. S5.
Transposon mutations downstream of the /11 integration site are transferred by lateral transduction. Next, we defined the type of transduction occurring with these 4 chromosomal transposon mutants. Lateral transduction occurs in bacterial DNA up to 100 kb downstream of a laterally transducing prophage based on the pac site encoded by that phage. In the case of f 11, the pac site was in the small terminase gene and this meant that the region downstream of the att R (,1.923 Mb) was transferred by lateral transfer while the region upstream of the att L (.1.967 Mb) was not (11,13). Therefore, we hypothesized that the downstream transposon mutations could be transduced laterally. To establish whether the higher transduction frequencies observed in the mutants with transposons located downstream of f 11 were due to the effects of the specific transposon insertions or the more general effect of lateral transduction, we investigated other transposon mutants from the same region of DNA. We used three different mutants where the transposon was located downstream of the prophage but had not been identified by the replicate 1 screen  Table S1 for NARSA equivalent), 'replicate' refers to which biological replicate screen the mutation was identified in, 'location' identifies whether the transposon insertion was in an MGE-situated gene or the chromosome (Chr). Upstream and downstream refer to the position of the gene relative to the f 11 prophage insertion, using the prophage directionality (eg. downstream = downstream of the att R , upstream = upstream of the att L ). Locations are color-coordinated with the genome map regions.
(Tn_1799, Tn_1843, and Tn_1852; Tn_1852 was subsequently identified in a later screen). We measured the transduction frequency of these transposon mutants in 8325-4 w 11 following induction. All three had a significantly higher transduction frequency than the Tn_rotB control ( Fig. 4) and were comparable to the Tn_1849 and Tn_1797 mutants.
Because these mutants had a high transduction frequency similar to that of the mutants from the screen, it suggested that lateral transduction was responsible. Because lateral transduction only occurred when a prophage was induced, we tested the transduction frequency of both the original mutants and these 'neighboring' mutants following infection where lateral transduction could not occur. The transposon mutants in the JE2 background were infected with f 11 and the transduction frequency of the resulting lysate was established. The transduction frequencies of all the transposon elements located downstream of f 11 were higher than the Tn_rotB control ( Fig. 5A and B). However, these transduction frequencies were significantly reduced for Tn_1843, Tn_1852, and Tn_1797 compared to the previous inductionlinked frequencies (Fig. 5C to E). Furthermore, Tn_1799 and Tn_1849 both showed lower transduction frequencies following infection, although these differences were not significant. This showed that lateral transduction was the main cause of the increased transduction frequencies downstream of the f 11 integration site.  Transposon mutations upstream of the /11 integration site are transferred by specialized transduction. To establish whether the genes located upstream and close to the f 11 integration site were transduced at higher frequencies due to gene-specific effects or because of specialized transduction, additional transposon mutants from this region were examined. The mutants in the same region as Tn_1866 and Tn_1868 showed comparable transduction frequencies, while the mutant with the transposon element furthest from the phage integration site (62 kb upstream) showed a gradual reduction in transduction frequency, albeit notably higher than the Tn_rotB control ( Fig. 6; Tn_1865, Tn_1883, and Tn_1898 versus Tn_1916, respectively).
Furthermore, when antibiotic resistances from the original JE2 transposon mutants were transduced following f 11 infection, the transduction frequencies were lower and not significantly different to the Tn_rotB control (Fig. 7). Given that specialized transduction did not occur during infection, these results suggested that specialized transduction was occurring in the region immediately upstream of the f 11 and caused the higher levels of transduction seen. Indeed, by using primers annealing within the NTML transposon (Martn-ermR) and the f 11 phage (phi11-direction-3f) it was possible to amplify products from induced lysates of 8325-4 f 11 Tn_1866 and 8325-4 f 11 Tn_1868 (Fig. 8). This indicated that some transducing particles in these lysates contained the upstream regions of bacterial DNA alongside the 59-end region of the phage genome, which would be expected for specialized transduction. Likewise, when we transduced the Tn_1866 and Tn_1868 transposon mutations into an 8325-4 as opposed to 8325-4 f 11 recipient, the percentage of 8325-4 recipients that by PCR were also positive for the Sa5 Int encoded by f 11 was significantly higher than observed when examining cells where the Tn_RotB had been transduced (Fig. S6). This would be expected with specialized transduction where the majority of transductant DNA injected into a new recipient cell would recombine with an integrated coinfecting phage and, therefore, selecting for a higher proportion of lysogenic recipient cells (18).  Regions identified from the preferentially transduced transposon elements are packaged at higher frequencies than the rest of the genome. To support our genetic screen, we sequenced phage f 11 lysates from mitomycin C induced 8325-4 f 11 lysogen to directly identify bacterial DNA packaged in phage capsids. Fig. 9 shows the mapping of the reads to 8325-4, thus removing reads mapping to the phage and allowing the lower frequency reads that mapped to the bacterial genome to be seen. Regions of elevated coverage were identified by dividing the genome into 13.5 Kbp bins and comparing the average coverage for each bin. The region between 1.09 and 2.01 Mbp was identified as the single region with significantly elevated coverage, with a mean depth overall of 5.6. This corresponded to the regions upstream and downstream of the f 11 att B (indicated by the red line), supporting our screening data in which the identified mutants were clustered in this same region. The rest of the genome showed only the baseline level of reads mapping, with an average mean depth overall of 1.1. The pattern of the downstream bacterial DNA indicates that lateral transduction may be responsible, with a gradual reduction of mapped reads from high numbers near to the att B , decreasing further away from the attachment site. Lateral transduction has been shown to taper off in this fashion, while still facilitating the transfer of ;200 kb of downstream bacterial DNA (10). Here, we see elevated read mapping as far away as 798 kb downstream from the att B site (mean depth 1093501 to 1107000 bp, 2.6). The read alignments of the upstream bacterial DNA peak encompass 117 kb of DNA upstream of the att B (mean depth 1998001 to 2011500 bp, 5.6). These reads could represent the effect of specialized transduction following the incorrect excision of the phage, albeit representing a larger region than one would expect with a phage genome length of ;45 kb. In addition, there was mismapping of phage reads forming a large peak left of the att B and a single large peak to the right of the att B mapped to the ribosomal operon containing three ribosomal genes encoding 5S, 23S, and 16S rRNA, these were excluded from the analysis. . N indicates no template. The expected size for region f 11 to Tn_1866 was ;10.6 kb and ;11.7 kb for region f 11 to Tn_1868. Three biological replicates were performed, including induction, precipitation, DNA extraction, and PCR, and the above gel is representative. Primers Martn-ermR and Phage direction-3f were used for all templates (Table S2).
Based on our observations that these upstream regions were being transduced alongside the phage genome (Fig. 8), together with the unexpectedly large regions of elevated read mapping of these upstream regions (Fig. 9), we hypothesized that the transduction of this region was more complex than the wrongful excision of the prophage. Traditionally specialized transduction is defined as misexcision where nonhomologous recombination of the incomplete phage and bacterial flank leads to their circularization and pac-based packaging. This would limit specialized transduction to the flanking regions less than a phage genome-length (;45 kb, plus redundancy) away from either the att L or att R . However, here we identified elevated reads mapping and increased transduction frequencies up to 117 kb upstream of the att B ( Fig. 6 and 9, respectively). Interestingly, outside the region of elevated reads, the transduction frequencies were not significantly different from the Tn_RotB control (Fig. S7).
We hypothesized that the extended regions of elevated transduction frequencies could be explained by homologous regions flanking the integrated prophage and upstream region enabling homology-based circularization of this region, including the phage. BLASTn of the nucleotide sequence up to 45 kb downstream of the att B with the upstream region of elevated reads mapping, indicated a 3.1 kb region of 99% identity between a region at the end of the elevated reads mapping and another ;18 kb downstream of the att B . If these regions were circularized it would include the complete prophage genome, which could facilitate packaging of both phage and bacterial DNA from the pac gene while integrated into this episome-like circular structure, similarly to the packaging of integrated DNA during lateral transduction. To confirm whether this circularization could occur, primers were designed to amplify the hypothetically circularized region. Following large plasmid extraction, we could amplify the product from both noninduced and induced 8325-4 f 11 (Fig. S8).

DISCUSSION
We described a high-throughput method by which the mobile regions of bacterial genomes can be identified. By screening pooled transposon mutants, we were able to where the genes referred to in this study are represented by blue boxes, with the NTML gene numbers included. The central two lines represent the forward and reverse DNA strands, with the concatenated assembly contigs identified with alternating orange and brown boxes. The narrow peak of read mapping to the left of the insertion site is caused by mismapping of phage reads, and the narrow peak to the right is caused by multiple mapping to an rRNA operon.
identify different modes of phage-related horizontal gene transfer, including lateral and specialized transduction. The method could be applied to any strain for which there is a transposon mutant library available. Before our screen, we anticipated obtaining two types of mutants; namely, those where the transposon was placed in a genome location transferred with higher frequency, for example, by lateral transduction and mutants that due to inactivation of a bacterial gene would lead to increased transduction efficacy. The results showed, however, that we primarily obtained mutants belonging to the first category and, in addition, transposon elements that had been inserted into MGEs. Thus, for three of the resulting transposon mutants from the first replicate screen, the transposon element was inserted in phages being endogenous to the JE2 strain, namely, the USA300f Sa2 and USA300f Sa3, and these had been transferred to the 8325-4 background. This phage transfer was not unexpected because prophages can be induced by an infecting phage and the 8325-4 strain encodes attachment sites for these phages (11,12,31).
For two of the transposon mutants preferentially selected in our first screen, the transposon elements were located downstream of the f 11 attachment site (SAUSA300_1849 and SAUSA300_1797). These mutants displayed high transduction frequencies following induction, compared to a control where the antibiotic resistance marker was located elsewhere in the bacterial genome, and this effect was reduced when the transfer was monitored following phage infection. We concluded that the increased transduction frequency for these mutations was dependent on the induction of an integrated f 11 and that they were transferred by lateral transduction (10). This notion was supported by bacterial DNA sequences obtained from induced phage lysates that indicated extensive packaging of bacterial DNA downstream of the integrated phage. Interestingly, these data implicate lateral transduction at an even greater distance downstream of the att B site than has previously been suggested (10), with slightly elevated reads mapping up to 798 kb downstream.
It has been shown that temperate phages are capable of both generalized and specialized transduction (16,32), but specialized transduction had not been reported in S. aureus until now. However, for the transposon mutants Tn_1866 and Tn_1868 identified in our screen, the transposon element was located immediately upstream of the f 11 integration site at a distance of 7.6 and 8.7 kb, respectively, thus placing it comfortably within a region where specialized transduction could be expected to occur. The increased transduction frequencies of Tn_1866 and Tn_1868 were eliminated when lysates were created by infection rather than induction, removing the possibility of generalized transduction. Equally, they would not be transferred by lateral transduction, which occurs downstream of the phage pac site (10). This left specialized transduction as the remaining, feasible explanation. Indeed, we were able to identify PCR products containing both f 11 and the transposon using precipitated phage lysate from the strains with Tn_1866 and Tn_1868. This indicated that some phage capsids contained both the att L region of the phage together with the upstream flanking region up to 8.7 kb away. Further, the majority (91% and 86%, respectively) of 8325-4 Tn_1866 and 8325-4 Tn_1868 transductant colonies were positive for the w 11 integrase. For Salmonella phage P22 it has been seen that specialized transduction into a recipient bacterium is mediated by recombination with an integrated prophage (15,18). Therefore, we expect that in our case, specialized transduction may involve recombination between a transducing particle carrying DNA upstream of the phage and a coinfecting phage during infection. These data, support that the upstream genes are transferring through specialized transduction.
The DNA sequencing of the precipitated phage particles from induced 8325-4 f 11 correlated well with the transduction frequencies of our selected transposon mutation strains. As such, we were initially confused by the extended region of elevated reads mapping to 117 kb upstream of the phage attachment site. While the transposon mutations in genes 1866 and 1868 sit well within the region of classical specialized transduction (less than 10 kb upstream of the att B ), the mutation in Tn_1916 is 62 kb upstream and still showed significantly higher transduction frequencies than the Tn_RotB control. Although this correlated with DNA sequencing, it could not be transferred by classical specialized transduction. Outside of the region indicated by our sequencing data, we no longer saw the higher transduction frequencies. Using the DNA sequencing data, we identified the end of the elevated reads mapping and found that this region had a 3.1 kb homologous region with 99% identity downstream of the phage. We hypothesized and confirmed that this region could recombine and in doing so form a large circle of DNA that would include the complete f 11 prophage. We speculate that the homologous regions up and downstream of the f 11 could facilitate a new variation of specialized transduction involving circularization of both prophage and flanking regions. Future work into this interesting phenomenon will look into the molecular mechanism by which such circularization could take place and how it is related to phage replication and packaging.
A limitation of our screen is the coselection for transposon mutants with greater fitness. This may explain why within our pooled mutants we identified a limited number of transposon mutations from the highly transduced regions, even though others were present in the pool. The transposon mutations identified in the screen may well represent those of higher fitness relative to the other highly transduced transposon mutants from that pool.
This study acts as a proof of concept for using transposon mutant libraries to investigate which regions of the bacterial genome are highly mobile. We showed phage transfer, generalized, specialized, and lateral transduction between staphylococcal strains. While the setup employed here was used to identify highly transduced genes and chromosomal locations, it may also reflect what happens over time in the real world. The release of f 11 from lysogens promotes the acquisition of bacterial DNA by auto-transduction and resulting transductants will, as lysogens, be able to repeat the process (33). Horizontal gene transfer underpins the evolution and adaptability of S. aureus and approaches such as the one adopted here will help us increase our understanding of which regions are transferred and how.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial strains used in this study are listed in Table S1. S. aureus strains were grown in tryptic soy broth (TSB) and on tryptic soy agar (TSA) plates (Oxoid) with or without antibiotics as appropriate.
Growth curves. Bacterial growth curves were performed in TSB using the bioscreen C automated microbiology growth curve analysis system. Briefly, overnight cultures were subcultured in TSB to an optical density at 600 nm of 0.01 and 250 mL plated in each well of the bioscreen C honeycomb 100-well plates. The OD 600 was recorded every 30 min for 24 h at 37°C with shaking, which was paused 5 s before each read. Growth curves were analyzed using the GrowthCurver R script and the t_gen (or doubling time) was compared between strains.
Bacteriophage transduction. Phage transductions were performed as previously described (10,33). Briefly, recipient strains were grown to 1.4 OD 600 , 4.4 mM CaCl 2 added, and 1 mL aliquots of recipient mixed with 100 mL phage lysate, at different dilutions in phage buffer (10 0 to 10 24 ). After 20 min incubation at 37°C, 3 mL of liquid TTA (transduction top agar; TSA, 50% agar) was added and the mix was poured out on selective TSA plates. Plates were incubated at 37°C for 24 h and colonies were counted.
The transduction frequency was calculated using the CFU/mL divided by the PFU/mL to control for variation between lysates.
Bacteriophage infection and induction. Bacteriophage infection was performed as previously described (33). Briefly, recipient bacteria were grown to 0.15 OD 600 and centrifuged, before resuspending the pellet in 1:1 TSB and phage buffer. The culture was then infected with phage at a multiplicity of infection (MOI) of 1 and incubated at 30°C, 80 rpm until complete visual lysis, normally after 3 to 4 h. Lysates were then filtered using 0.22 mM filters and stored at 4°C.
Bacteriophage induction was performed as previously described (10,33). Briefly, lysogens were grown to an OD 600 of 0.1 to 0.2 and 2 mg/mL mitomycin C (Sigma, from Streptomyces caespitosus) was added to induce resident prophages. Cultures were then incubated at 30°C, 80 rpm until complete visual lysis. Lysates were filtered using 0.22 mM filters and stored at 4°C.
NTML transposon transduction screening. One 96-well plate was replicated and grown overnight in selective TSB (erythromycin, 5 mg/mL) before pooling 10 mL of each mutant in 100 mL selective TSB in a 500 mL flask. The culture was infected with f 11 (MOI 1) as described. The lysate was then used to transduce the NTML transposons to 8325-4 f 1, as described. The transduction plate of undiluted lysate was harvested, pooled, and grown in selective TSB and the culture was diluted then induced. This lysate was used to transduce into the 8325-4 f 11 recipient a second time, and the transductants were pooled and induced. From the resulting lysate, a third transduction was performed, 10 transductant colonies selected, or all colonies were pooled, and the location of the transposon mutation(s) identified as previously described (28,34).
DNA sequencing. Preparation of the phage lysates for DNA sequencing was done by precipitating the phage capsids before DNA extraction. For the phage precipitation, DNase (2.5 U/mL, Thermo) and RNase (1 mg/mL, Sigma) were added to the filtered lysate before incubation at 37°C for 1 h. NaCl 2 (1 M) was added and mixed on the ice at 4°C for 1 h. The lysate was centrifuged at 11,000 Â g for 10 min at 4°C and the supernatant was collected, mixed with PEG 8000 (10% wt/vol), before incubation on ice overnight at 4°C. The lysate was centrifuged at 11,000 Â g for 10 min at 4°C and supernatant discarded, with falcon tubes dried. The pellet appeared as two vertical white lines and was resuspended in 8 mL phage buffer per 500 mL lysate, with resuspension for 1 h at 4°C. DNA extraction was performed using the MagAttract HMW DNA kit (Qiagen) per the manufacturer's instructions. DNA concentration and absorbance peaks were checked using a Thermo NanoDrop 2000 and a Qubit Flex Fluorometer. A draft genome for 8325-4 was generated by sequencing chromosomal DNA on an Illumina miSeq following the manufacturer's instructions. The sequence was assembled, and the contigs were ordered using the Sanger Institute pipelines (35). Phage lysate DNA was sequenced in the same way, and the resulting reads were mapped to this genome using SMALT (35,36) and visualized using Artemis (35,37).
To identify regions of elevated coverage, the genome was split into 198 bins of 13.5 Kbp, and an area of elevated coverage was identified between 1093501 and 2011500 bp, corresponding to bins 82 to 149, after calculating the coverage for each bin. To test whether this region had significantly elevated coverage compared to the rest of the genome, an independent t test was performed comparing the coverage observed in these two bin categories. Further statistical analyses were performed to test whether the elevated coverage at the ends of the elevated region remained above the background coverage of the rest of the genome. A one-sample t test was performed to compare the mean of the coverage distribution of bins 1 to 81 against bin 82, and another to compare bin 149 with the mean of bins 150 to 198.
DNA methods. General DNA manipulations were performed using standard procedures. Primers used in this study can be found in Table S2. PCRs were performed using DreamTaq Green master mix (2Â; Thermo Fisher Scientific) per the manufacturer's instructions. PCRs for identification of transposon locations were performed as previously described (28). PCR products were sequenced at Eurofins MWG Operon.
For identification of prophages in strains, a multiplex PCR for the main integrases of the staphylococcal phages was performed, as previously described (30). Initial screening of colonies was performed using 1 mL of template from a single colony resuspended in 50 mL ddH 2 O, boiled for 10 min, and on ice for 10 min. Results were confirmed using DNA template (10 ng), extracted using GenElute TM Bacterial Genomic DNA kit as per manufacturer's instructions (Sigma-Aldrich). For integrase checks of transductants, colony PCRs using just the Sa5 Int primers was performed as described above.
For long amplifications, the Takara LA polymerase PCR kit was used per the manufacturer's instructions with the following alteration, a 30 s annealing step was added at 55°C. Templates used were either bacterial DNA extracted with the GenElute TM Bacterial Genomic DNA kit or phage DNA extracted from phage lysates that had been precipitated, as previously described, and extracted using the GenElute TM Bacterial Genomic DNA kit with an extended proteinase K incubation of 1 h (Sigma-Aldrich). Fifty nanograms of the template were used for each PCR. The resulting products were run out on a 0.7% agarose gel at 30 V overnight.
For extraction of the large plasmid-like structures, samples were grown for 1 h, before induction with MitC as described. Samples were taken at the time of induction (T0) and 1 h after induction, with noninduced samples also taken. Fifty milliliters of the culture was taken per sample and prepared as previously described (38), with the amounts doubled and RNase added to the resuspension buffer. Sixty nanograms of the template were used for subsequent PCRs using a DreamsTaq polymerase kit.
Statistical analyses. Data analyses were performed in GraphPad Prism 8 software. Specific statistical tests are indicated in the figure legends, where appropriate.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1.5 MB.