Characterization of the genetic switch from phage ɸ13 important for Staphylococcus aureus colonization in humans

Abstract Temperate phages are bacterial viruses that after infection either reside integrated into a bacterial genome as prophages forming lysogens or multiply in a lytic lifecycle. The decision between lifestyles is determined by a switch involving a phage‐encoded repressor, CI, and a promoter region from which lytic and lysogenic genes are divergently transcribed. Here, we investigate the switch of phage ɸ13 from the human pathogen Staphylococcus aureus. ɸ13 encodes several virulence factors and is prevalent in S. aureus strains colonizing humans. We show that the ɸ13 switch harbors a cI gene, a predicted mor (modulator of repression) gene, and three high‐affinity operator sites binding CI. To quantify the decision between lytic and lysogenic lifestyle, we introduced reporter plasmids that carry the 1.3 kb switch region from ɸ13 with the lytic promoter fused to lacZ into S. aureus and Bacillus subtilis. Analysis of β‐galactosidase expression indicated that decision frequency is independent of host factors. The white “lysogenic” phenotype, which relies on the expression of cI, could be switched to a stable blue “lytic” phenotype by DNA damaging agents. We have characterized lifestyle decisions of phage ɸ13, and our approach may be applied to other temperate phages encoding virulence factors in S. aureus.

1991; Xia & Wolz, 2014). As the beta-hemolysin encoded by hlb is also a virulence factor (Huseby et al., 2007;Katayama et al., 2013), this negative conversion appears counterintuitive, but Sa3int phages have been shown to excise and remain as pseudo-lysogens, allowing expression of both the beta-hemolysin and the virulence factors encoded by the phage Katayama et al., 2013;Salgado-Pabón et al., 2014).
Temperate phages have a dual lifecycle involving either lytic replication and production of active phages or being inserted in and replicated with the bacterial chromosome. The decision between these lifestyles is made by a phage-encoded (epi)genetic switch. Phage λ, targeting Escherichia coli, offers the classical example of such a switch, where a complex decision phase determines from which of the two involved promoters, P R or P RM , transcription occurs.
Expression from the P RM promoter produces the phage repressor CI, whereas expression from P R allows the production of the Cro repressor, as well as other gene products needed for producing active phage particles (Casjens & Hendrix, 2015;Oppenheim et al., 2005).
Stable expression from the P RM promoter is needed for the maintenance of the λ chromosome into the E. coli chromosome. It also produces a high concentration of the CI repressor, which is required for activation of the P RM promoter and tight repression of the P R promoter in the integrated prophage and any incoming λ phages, rendering the lysogenic bacterium immune to secondary attacks. The decision and integration of phage λ are not solely dependent upon the phage-encoded factors, as E. coli host factors also affect this process (Grodzicker et al., 1972;Herman et al., 1993;Kihara et al., 2001;Roucourt & Lavigne, 2009). The escape of phage λ from the lysogenic state can be initiated by autocleavage of the CI repressor.
In a process analogous to autocleavage of the LexA repressor, CI is bound to activated RecA filaments polymerized on single-stranded DNA during the SOS DNA damage response, leading to activation of CI auto-peptidase activity (Atsumi & Little, 2006;Little, 1984;Oppenheim et al., 2005). A quite different switch mechanism is used by the TP901-1 phage infecting the Gram-positive bacterium Lactococcus lactis (Madsen et al., 1999) which shares extensive homology to the ɸ13 switch (Pedersen et al., 2020). Here, a CI repressor, expressed from the lysogenic promoter P R , represses the lytic P L promoter. A small repressor (MOR, for modulator of repression) expressed from the P L promoter represses the P R promoter; however, it does require CI as a co-repressor. In addition MOR functions as an antirepressor against CI (Pedersen & Hammer, 2008).
As a member of the staphylococcal Sa3int phage family, the bacteriophage ɸ13 shares features with the other members. These features include the Sa3 type integrase and the presence of the sak, chp, and scn virulence genes (Goerke et al., 2009;Xia & Wolz, 2014), but an analysis of the genetic switches within the group has never been reported.
To characterize the switch from phage ɸ13, we have compared its switch region to the switch regions in TP901-1 and the switch regions in Staphylococcus phages, showing that a group of phages with the Sa3 integrase shares similar switch regions containing CI and MOR homologs. We have purified the ɸ13 CI repressor and shown that it interacts with three palindromic sites in the switch region. By constructing plasmids containing the ɸ13 genetic switch with lacZ reporter fusions to the lytic promoter (switch plasmids) and analysis of the frequency of Lac phenotypes in transformants of the natural and a heterologous host, we have shown that a functional mor gene is required for "decision switching" and that this process is independent of staphylococcal host factors. The concept of "toggle switching" has previously been defined as "the induced or accidental change from one stable phenotype to another from a bistable switch" (Gardner et al., 2000). Thus, toggle switching is when the phage changes its state, either spontaneously or following induction. In contrast, decision switching is when the phage initially establishes itself as lytic or lysogenic, in a process that precedes toggle switching. In the current study, ɸ13 switch plasmids were found to toggle spontaneously from the "lytic" (Lac+, blue) colony phenotype to a stable "lysogenic" (Lac−, white) colony phenotype. In the opposite direction, toggling from the stable "lysogenic" colony phenotype to the "lytic" colony phenotype could readily be induced by sub-lethal concentrations of DNA damaging agents. The results from this study can prove important in the analysis of S. aureus virulence in vivo by determining conditions during infection that would result in induction and spread of ɸ13 and the unwanted establishment of ɸ13 prophages in susceptible hosts. The results may ultimately aid in the understanding of how some Sa3int prophages are disseminated in society and suggest ways to limit S. aureus colonization in humans.

| Growth specifications
Unless otherwise specified, E. coli and Bacillus subtilis were grown in lysogeny broth (LB) or LB agar, while S. aureus was grown in tryptic soy broth (TSB) or agar (TSA). All strains were grown at 37°C with sufficient aeration and the addition of appropriate antibiotics. 5-Br omo-4-chloro-3-indolylβd-galactopyranoside (X-gal) was added to agar plates at 100-200 µg ml −1 .

| Bacterial strains and plasmids
All bacterial strains used in this study can be found in Table A1, while selected strains, plasmids, and phages are summarized in Tables 1-3, respectively.

| Protein expression and purification
A synthetic codon-optimized gene for recombinant CI repressor protein with an N-terminal His-tag and TEV cleavage site (ɸ13 rCI) was purchased from Genscript ( Figure A1) in a pET30a(+) expression vector. The expression vector was transformed into BL21 (DE3) E. coli cells and grown at 37°C. ɸ13 rCI expression was induced by 0.5 mM isopropyl βd-1-thiogalactopyranoside in a mid-exponential culture and the temperature lowered to 25°C.
For EMSA using native CI, the tag on ɸ13 rCI was removed beforehand by mixing the protein 10:1 with TEV protease (Sigma Aldrich) and incubating it at 4°C for 3 days. The uncleaved protein was separated from the mixture by IMAC and the flowthrough was collected for use.

| DSF and CD spectroscopy
The inflection temperature, T i , was determined using a Nanotemper Tycho™ NT.6 DSF monitoring changes in the ratio of intrinsic fluorescence at 350 and 330 nm while heating from 35°C to 95°C with a temperature ramp of 30°C min −1 . The sample was loaded in high precision glass capillaries with a concentration of 4.9 µM.
For CD spectroscopy, the purified protein was transferred to a 20 mM NaF buffer, pH 7.5. The experiments were conducted on a Jasco J-815 CD spectropolarimeter in a suprasil quartz cell with a 1 mm light path (Hellma Analytics) and a sample concentration of 1.56 µM. The spectra were measured at room temperature in a continuous mode at 20 nm min −1 from 260 to 190 nm with a data pitch and bandwidth of 1 nm, averaging five spectra per measurement and subtracting the buffer spectra. The CD spectrum was deconvoluted and analyzed using DichroWeb (Whitmore & Wallace, 2008) with the CDSSTR method (Sreerama & Woody, 2000) and reference sets 3 and 6 and by BeStSel (Micsonai et al., 2015(Micsonai et al., , 2018. TL software (Amersham Biosciences). The image was analyzed using Fiji (Schindelin et al., 2012). The integrated density of each band was measured and used to calculate the individual fractions of protein bound to DNA at each protein concentration. The data were fitted using the non-linear curve fitting option and the specific binding with Hill slope option in GraphPad Prism (GraphPad Software, Inc.).

| Construction of switch plasmids
All PCRs were conducted at appropriate annealing temperature and elongation time using Phusion polymerase unless otherwise speci-

| Transformations
Electrocompetent E. coli cells were created and electroporated as previously described (Pedersen et al., 2020). Cells were plated on selective agar plates after 1 h of recovery.
Competent B. subtilis cells were obtained as described in (Konkol et al., 2013) with one minor change; 30 min after plasmid addition equal volumes of LB media were added to the transformation mixture, followed by incubation for 1 h and plating on selective plates.
Transformations in S. aureus were performed using the procedure for electroporation described in (Monk et al., 2015). pYL112Δ19 was transformed into 8325-4 and 8325-4 ɸ13-kana before the introduction of integrating plasmids to supply the required integrase.

| Transductions
Transductions in B. subtilis and S. aureus were accomplished by using phage SPP1 or ɸ11/ɸ80α, respectively. See Table 3

| Stability of switch plasmid phenotypes
To examine the stability of the "lysogenic" and "lytic" phenotype of the switch plasmids, a single colony resulting from transformation was suspended in an appropriate medium, diluted, and plated at a density allowing the formation of single colonies for phenotype determination.

| Plate inductions
To test the effect of inducing substances, both B. subtilis and S. aureus cells containing switch plasmids were plated at a cell density resulting in around 1000 colonies per plate, obtained by suspending a colony in LB/TSA medium, dilutions, and plating; 3 µl inducing substance (0.5 mg ml −1 mitomycin C, 1 mg ml −1 ciprofloxacin, or 30% hydrogen peroxide) was applied to the middle of the plate, followed by incubation at 37°C. When relevant, selected colonies were suspended in media, diluted, and plated at a density resulting in single colonies to determine phenotypes.

| Temporal induction
A culture of 8325-4 geh::pCSK9 favoring the "lysogenic" phenotype (CSK48) was inoculated in TSB and grown to OD 600 = 0.6 followed by induction by 1 µg ml −1 mitomycin C. Samples were taken periodically, diluted in TSB, and plated on selective TSA plates containing X-gal. Switch plasmid phenotypes were determined after overnight incubation.

| Phage infection of 8325-4 harboring switch plasmids
Cultures were inoculated in TSB and grown to OD 600 = 0.6. At this point, a sample was plated for determination of colony forming units (CFU) ml −1 , after which, phages were added at a MOI of 0.1, followed by incubation for 2 h. Cultures were spun at 5000 g for 5 min and the supernatant filter-sterilized (0.2 µm), followed by dilution in SM buffer, titration on bacterial lawns of 8325-4 (TSB, 0.5% agarose, 5 mM CaCl 2 ), and overnight incubation for determination of plaque forming units (PFU) ml −1 . The amount of phages produced per host cell was determined by calculating PFU ml −1 per CFU ml −1 .

| Lysogenization frequency of ɸ13
Exponential cultures of 8435-4 were infected with ɸ13-kana at a MOI of 0.1 and 1. Phages were allowed to adsorb for 15 min on ice, after which cells were centrifuged and the pellet was resuspended in SM buffer. Dilutions were plated on TSA plates containing kanamycin for the determination of lysogens. Dilutions were further titrated on bacterial lawns of 8435-4 to detect cells in which the phage favored the lytic lifecycle. Phage titer was determined as PFU ml −1 after overnight incubation. The lysogenization frequency was calculated as the ratio of lysogenic cells (CFU (Kan R ) ml −1 ) of all infected cells (CFU (Kan R ) ml −1 + PFU ml −1 ).

| Topology of the ɸ13 switch region and detection of identical switches in members of the Sa3int and the PVL encoding phages
Recently, a global search for DNA sequences similar to the phage TP901-1 switch region led us to identify a putative switch region in the S. aureus phage ɸ13 (Pedersen et al., 2020). As shown in the map in Figure 1 and the amino acid sequence alignments in Figure  The limited similarity was found to the C-terminal domain CI-CTD 1 from TP901-1, whereas the CI-CTD 2 showed no sign of conservation ( Figure A2). Instead of the CI-CTD 2 domain, the CI repressor from ɸ13 carries a region with clear similarity to the C-terminal domain of the CI repressor from phage λ (see Figure A2). The two catalytic residues from the peptidase domain as well as the autocatalytic site from λ are conserved in ɸ13, suggesting that the CI repressor from ɸ13, in contrast to the CI repressor from TP901-1 (Madsen et al., 1999), could be capable of autocleavage.
To analyze if the ɸ13 switch type was widespread among phages, we performed a homology search (Altschul et al., 1990) with the 1260 bp ɸ13 DNA sequence covering the region shown in Figure 1, excluding staphylococcal genome sequences. Only sequences from completed genomes of staphylococcal phages were identified, and six phages had hits with full coverage and sequence identity (>99% identity): PVL, tp301-1, P630, 3AJ-2017, IME1346_01, and ɸ13.
Two of these, 3AJ-2017 and IME1346_01, encode the virulence proteins Sac, Chp, and Scn (Oliveira et al., 2019) and belong to the Sa3int group like ɸ13, while the phages PVL and tp301-1 encoded the LukS-PV and LukF-PV PVL virulence factors (Oliveira et al., 2019). This shows that the homology between the switches from ɸ13 and TP901-1 is not unique and that the ɸ13 switch type determines the decisions between lysogenic and lytic growth for two important groups of phages carrying pathogenicity genes.

| ɸ13 CI binds to operator sites in the intergenic region between cI and mor
To characterize the switch present in ɸ13, it was first important to know that the postulated CI repressor was functional and had an affinity towards the postulated CI operators. To determine whether the CI repressor from ɸ13 binds to these putative operators, CI was expressed with an N-terminal His-TEV-tag (termed rCI, see Figure   A1) and purified by affinity chromatography and gel filtration ( Figure   A3a,b). After circular dichroism spectroscopy and differential scanning fluorimetry had confirmed that the protein was folded and stable ( Figure A3c,d), electrophoretic mobility shift assays (EMSA) were F I G U R E 1 Comparison of switch regions from S. aureus phage ɸ13, L. lactis phage TP901-1, and E. coli phage λ. The geometry of the switch region is not shown to scale. The mor gene of TP901-1 and ɸ13 is shown as a red box, the cro gene of λ is shown as a purple box, and the first part of the cI gene encoding the DNA binding N-terminal domain (CI-NTD) is shown as a blue box. The part of the cI gene from ɸ13 encoding the extreme C-terminal domain (CI-CTD 2 ) is not homologous to the CI-CTD 2 domain in TP901-1 but to the C-terminal domain from CI of λ, containing a peptidase domain. CI-CTD 2 is shown as a green box for TP901-1 and a yellow box for ɸ13 and λ. A dimerization domain (CI-CTD 1 ) with helical hook structures in TP901-1 connects CI-CTD 2 and CI-NTD and is shown as a black box for TP901-1 and ɸ13, while the dimerization region is shown as a gray box for λ. The divergently oriented promoters, P R and P L for TP901-1, P RM and P R for λ, and putative P R and P L promoters for ɸ13, are shown as arrows preceded by small black boxes symbolizing −10 and −35 regions.  Binding of rCI to these mutated control probes was severely weakened and no bands for specific binding could be detected ( Figure 2b).
Interestingly, it can be seen from Figure 2b, that the bands with CI-bound DNA probe (marked with red arrows) disappear at higher concentrations of CI. We hypothesize that this is due to multimerization of CI because the labeled DNA probe migrates slowly as a smeared band at very high CI concentrations (Lane 1 in all gels in

| A minimal 1.3 kb switch region from ɸ13 is capable of decision switching in S. aureus
After having shown that the cI encoded repressor can bind to an operator (O L ) overlapping the putative P L promoter as in the TP901-1 switch, the isolation of a functional ɸ13 switch for in vivo analysis of switch frequencies appeared to be feasible. Previously a DNA fragment spanning the cI and mor genes from TP901-1 had successfully been inserted into a promoter fusion plasmid (Madsen et al., 1999), and used for detailed characterization of the TP901-1 switch (Pedersen & Hammer, 2008) through transformation of its L. lactis host. To monitor the regulatory decisions upon introduction of ɸ13 switch DNA, we constructed plasmids with transcriptional fusions to the P L promoter of the ɸ13 switch region, as illustrated in Figure 3, with a lacZ reporter gene located after the mor gene. Upon transformation and plating on selective plates containing X-gal, the resulting colonies were either blue, indicative of high mor expression from the lytic P L promoter, referred to as the "lytic" phenotype, or white, indicative of low mor expression due to repression of the P L promoter by CI expressed from the lysogenic P R promoter, referred to as the "lysogenic" phenotype.
Plasmid pCSK3 replicates in S. aureus and upon transformation of strain 8325-4 with pCSK3 we observed an average of 4.8% transformants with the "lysogenic" colony phenotype and 95% with the "lytic" phenotype (Table 4). This result indicated that the switch region present in pCSK3 could indeed perform decision switching in S. aureus. The genetic integrity of the switch region was verified by purification and subsequent re-transformation of pCSK3 isolated from 8325-4/pCSK3 transformants, yielding similar decision frequencies (see footnote to Table 4). Since the decision analysis was performed using a plasmid and not an intact phage, the use of pCSK3 as a model system for the ɸ13 decision switching required that the frequency of the "lysogenic" phenotype reflected the true lysogenization frequency. Infection of strain 8325-4 with the ɸ13kana phage was therefore performed at a MOI of 1 and 0.1 at a bacterial concentration of 1.3 × 10 8 CFU ml −1 . As shown in Table   A3, the average lysogenization frequency from five replicates was 4.6% ± 2.8% and 3.5% ± 3.3% for infections at an MOI of 1, and 0.1, respectively. Since the pCSK3 "lysogenic" (4.9%) and ɸ13 lysogenic frequencies are not significantly different (p = 0.87 and 0.47, for the two MOIs, respectively), we believe that the pCSK3 switch plasmid is a faithful biological model for ɸ13 switching. Finally, to ensure the importance of the MOR protein for the lytic pathway of the ɸ13 decision process, the start-codon in the mor gene was changed to a stop codon. The resulting plasmids, pCSK12, and pCSK13 acquired the "lysogenic" phenotype in 100% of transformations when introduced into S. aureus 8325-4, showing that the "lytic" phenotype is dependent upon the MOR protein (Table A4).
To analyze whether the decision frequency of the ɸ13 switch was dependent upon replication of the plasmid containing the switch DNA during the decision process, we introduced the switch on an integrative plasmid, pCSK9. Plasmid integration occurs at the bacterial attachment site of the phage L54a located in the lipase gene geh of S. aureus and is catalyzed by the phage integrase provided by plasmid pYL112-19 (Luong & Lee, 2007). Transformation of S. aureus 8325-4 with pCSK9 yielded a higher number of "lysogenic" colonies than with pCSK3 (12% compared to 4.9%, see Table 4). However, the very low efficiency of transformation (in total 77 transformants from 5 individual transformations) resulted in a high degree of variation, preventing a reliable determination of the decision frequency.
When a liquid culture of the "lysogenic" 8325-4/pCSK3 (CSK33) was infected with ɸ13-kana, the "lysogenic" phenotype was found to restrict the production of phages (compare "B" and "E" in Figure 4). The same result was found when a transformant carrying the integrated pCSK9 plasmid (8325-4 geh::pCSK9, CSK48) was inoculated from a "lysogenic" colony and was infected with ɸ13-kana (compare "A" and "C" in Figure 4), showing that CI expressed from a chromosomally integrated ɸ13 switch conferred immunity to attacking ɸ13 phages.
When cells with the integrated switch plasmid were inoculated from a "lytic" colony and infected with ɸ13, the presence of the switch did not affect phage production (compare "A" and "D" in Figure 4).
Surprisingly, however, when cells containing the replicative plasmid pCSK3 were inoculated from a "lytic" colony and infected by ɸ13-kana, the "lytic" switch phenotype inhibited phage production (compare "B" and "F" in Figure 4). It has previously been shown that replicating switch plasmids of TP901-1 favoring P L ("lytic" phenotype) resulted in partial immunity towards infecting phages (Madsen et al., 1999), in accordance with the present results. This suggests that, although there is not enough CI repressor present to repress P L on the multi-copy switch plasmid, the concentration is still sufficient to prevent lytic phage development. Upon screening of kanamycin resistance in cells surviving ɸ13-kana infection, very few lysogenic cells were detected, showing that survival under these conditions is not due to the formation of ɸ13-kana lysogens. Thus, the multicopy plasmid-derived "lytic" switch phenotype appears to inactivate phage development in either direction.

| Decision switching by the minimal switch region from ɸ13 does not require S. aureus host factors
As E. coli host factors are known to affect the integration and induction of phage λ (Roucourt & Lavigne, 2009), it was of interest to establish whether the ɸ13 switch can function independently of host factors. Switch plasmid pCSK3 has an origin of replication that is functional in most Gram-positive bacteria (Kovács et al., 2010), and we could therefore also examine ɸ13 switching in F I G U R E 3 Reporter plasmids for monitoring switching of the ɸ13 switch in B. subtilis and S. aureus. The genetic switch of ɸ13 was inserted into the reporter plasmids pDG1729 (pCSK2), pNZlac (pCSK3), and pCL25 (pCSK9) observed an average frequency of "lysogenic" colonies of 2%-3% (Table 4), which is close to the frequency obtained in S. aureus.
This suggested that the decision frequency was not affected by specific host factors.

| Low spontaneous induction frequency from the "lysogenic" phenotype to the "lytic" phenotype
To examine the stability of the ɸ13 "lysogenic" colony switch phenotype, simulating the stability of ɸ13 lysogenic cells, we suspended and plated a large number of cells from "lysogenic" colonies on agar plates containing X-gal. Around 0.2% "lytic" colonies were detected in otherwise "lysogenic" colonies from both B. subtilis and S. aureus. This low level of toggling from the "lysogenic" to the "lytic" colony phenotype is hypothesized to be due to spontaneous induction of the SOS response, which is inevitable and leads to TA B L E 4 Frequencies of colonies with "lysogenic" phenotype after the introduction of ɸ13 switch plasmids into S. aureus 3.8% ± 1.6% (n = 10) 3.0% ± 1.6% (n = 18) Note: Switch plasmids were replicative (pCSK3) or required integration into the host (pCSK9 in S. aureus, pCSK2 in B. subtilis). Recipient bacteria were naïve (S. aureus 8325-4 or B. subtilis 168) or ɸ13 lysogens (S. aureus 8325-4 ɸ13-kana). The number of biological replicates with more than 100 resulting colonies is stated in brackets. Frequencies of "lysogenic" transformant colonies on plates containing X-gal were calculated as the frequency between "lytic" and "lysogenic" in each transformation event and the average is shown here. Standard deviations are based on differences between each event.
b Indicates <100 transformants obtained in each experiment. The number of "lytic" and "lysogenic" colonies obtained in total for all replicas followed by the total count of biological replicates (n) for each event are stated in brackets.
c Re-transformation of 8325-4 by pCSK3 plasmids extracted from "lytic" and "lysogenic" colonies of 8325-4 ɸ13-kana/pCSK3 yielded frequencies of "lysogenic" transformant colonies of 3% and 11%, respectively. When these plasmids were used to transform 8325-4 ɸ13-kana, 100% "lysogenic" transformants were obtained. prophage induction in bacteria that contain an SOS response and an intact recA gene . To distinguish the process of toggle switching from a mutational event leading to the "lytic" colony phenotype, we purified pCSK3 plasmid from "lytic" spontaneous revertants and re-transformed them into the same host, with and without the ɸ13 prophage. After transformation of the strains with the extracted pCSK3 plasmid, we obtained a frequency of "lysogenic" transformant colonies at 5% for re-transformation of 8325-4, and at 100% for re-transformation 8325-4 ɸ13 kana, indistinguishable from frequencies obtained by transformation with the original plasmid, and showing that the change was epigenetic rather than mutational.

F I G U R E 4 ɸ13-kana infection of
3.7 | High toggle frequency of the ɸ13 switch from the "lytic" to the "lysogenic" phenotype While the "lysogenic" colony phenotype was very stable, the "lytic" colony phenotype was subject to frequent toggling to the "lysogenic" phenotype. When the cells from small (young) "lytic" colonies were suspended in medium and plated on agar plates containing Xgal, around 3% of the colonies had assumed the "lysogenic" phenotype. Plating of larger (older) "lytic" colonies resulted in an average of 60% "lysogenic" colonies. Because of the somewhat artificial nature of the "lytic" phenotype, the stability of the phenotype was not analyzed further.

| Induction of the "lytic" switch phenotype by DNA damaging agents
Above, we reported that 0.2% of cells that had grown in "lysogenic" colonies on an agar plate had spontaneously toggled to the "lytic" phenotype, possibly initiated by the spontaneous induction of the SOS response. A more dramatic visualization of the interplay between the DNA damage and the ɸ13 switch could be observed by exposing "lysogenic" cells to a gradient of DNA damaging agents on agar plates containing X-gal. Induction of "lysogenic" (white) to "lytic" (blue) toggle switching is intimately connected to the level of the DNA damaging compound as seen in Figure 5. In our working hypothesis, the cellular CI concentration is determined through a race between CI expression and CI degradation, where a threshold CI concentration has to be crossed if the switch should toggle from the "lysogenic" to the "lytic" phenotype. Two clear zones can be de- peroxide are all known to induce ɸ13 lysogens (Goerke, Koller, et al., 2006;Tang et al., 2017), and Figure 5 shows the response of the ɸ13 switch plasmid to gradients of DNA damaging agents in its natural host ( Figure  Across the toggle-inducing zone of mitomycin C and ciprofloxacin the intensity of blue color decreases (top four panels in Figure 5), likely due to smaller fractions of toggle-switched cells in the colonies. To quantify the fractions of stable toggle-switched cells, we picked colonies grown at known distances from an application spot containing mitomycin C and determined the fraction of cells in each colony that could form colonies with a stable "lytic" phenotype. After a suspension of each colony in a fresh medium, plating, incubation, and quantifying the fraction of "lytic" colonies, an exponential dependency was observed between the frequency of "lytic" cells and the distance to the application spot for both S. aureus and B. subtilis ( Figure 6). This could suggest a quasi-linear dependence between the toggle switch frequency and the mitomycin C concentration since we expect a similar decrease in concentration by diffusion of a compound in agar (Koch, 1999), although the inherent instability of the "lytic" phenotype detected above should be taken into consideration. This dependence of toggle frequency on the concentration of the DNA damaging agent suggests the existence of high stochasticity in either the level of DNA damage, the concentration of RecA* nucleofilaments, the rate of autocleavage of CI, or the competition F I G U R E 6 Toggle switching of "lysogenic" cells in response to mitomycin C. Bacteria from a "lysogenic" (white) transformant of (a) 8325-4 geh::pCSK9 (CSK48) and (b) B. subtilis 168 thrC::pCSK2 (CSK9) were plated, and a small drop of mitomycin C was immediately placed in the center. After growth overnight, individual colonies at different distances to the mitomycin C application spots were picked and suspended in medium without mitomycin C, from which dilutions were plated on LB/TSA X-gal plates. The frequency of colonies with the "lytic" (blue) phenotype was determined and plotted against the distance to the mitomycin C application. Lines have been inserted to illustrate the relationship between colonies on the plates and the determined frequencies of toggle switching between the elements in the ɸ13 switch. Experiments to unravel the kinetic relationship between induction of the SOS response and the toggle switching of the genetic switch would be interesting, but it would require simultaneous monitoring of the SOS response.

| Timing of commitment to the "lytic" phenotype in S. aureus
To examine the timing of the response of the switch to DNA damage, we exposed "lysogenic" 8325-4 geh::pCSK9 (CSK48) cells to mitomycin C at a final concentration of 1 µg ml −1 . Induction of toggle switch to the "lytic" phenotype was monitored by plating dilutions of the mitomycin C treated culture on TSA plates containing X-gal. A "lytic" phenotype indicated that the induced cell that formed the colony was committed to the "lytic" phenotype and had passed on the phenotype to the descendants in the colony. Figure 7 shows the temporal increase in the fraction of cells committed to forming colonies with a stable "lytic" phenotype (in percent). The highest frequency of commitment to the "lytic" phenotype was obtained between 5 and 40 min of exposure to mitomycin C, and as many as 30% of all cells became dedicated in 20-30 min. However, because of the stochastic nature of the induction process, a fraction of the bacteria showed commitment in <5 min of exposure to mitomycin C (3% of the cells), while other cells required exposure for more than 40 min before commitment (10% of the cells). As discussed above, the CI threshold allowing toggle switching is determined by many factors, such as the extent of DNA damage, the CI degradation rate, the CI synthesis rate, and ɸ13 switch kinetics, each adding to the overall stochasticity of the process.

| CON CLUS IONS
In this study, we have reported the first analysis of a genetic switch from a Sa3int bacteriophage with implications for the phage-derived host specificity of S. aureus bacteria towards humans. As identical switches are present in other Sa3int phages and unrelated phages carrying genes for PVL, the results are also relevant for a larger group of pathogenic phages. Inhibition of the process leading to phage integration into S. aureus hosts could influence their ability to colonize humans. Therefore, it is crucial to understand the process by which the phage switch decides between (i) lysogenization of the host to integrate as a prophage or (ii) multiplying and killing of the host. It is equally important to understand the toggle process that induces a Sa3int prophage to kill the bacterium and release a litter of newly produced phages. We have shown that a minimal genetic switch of ɸ13, contained on a 1.3 kb DNA fragment and expressing functional cI and/or mor genes from divergent promoters, is fully competent in both decision switching and toggle switching. Purified his-tagged CI protein was found to be folded, stable, and to bind with high affinity to operators with two half-sites following the AGTTCAYR consensus, but with low affinity to mutant operators (AGTTGAYR). The two alternative phenotypes of the epigenetic switch are not equally stable. While the "lysogenic" phenotype is very stable (although chosen in <5% of the decision events) and leads to the lysogenic life cycle of the phage, the frequent "lytic" phenotype that resembles the first phase in the lytic life cycle is unstable and toggles towards the "lysogenic" phenotype. If this high toggle frequency indicates that prophage induction may be reversible in its early phase, then our preliminary findings that the nutritional status of the cell can modulate the toggle frequency could prove relevant.
Induction of "lysogenic" to "lytic" toggling by DNA damaging agents, resembling the induction of the ɸ13 prophage, was visualized in a plate diffusion assay that enabled us to locate MIC and MTC values ( Figure 5). In this work, we defined MTC values as a tool to distinguish between the antibiotic killing of cells by inhibition of cell function from killing by induction of phage lysis. Relative MIC/MTC ratios could be important in the clinical use of DNA damaging antibiotics to prevent phage lysis mediated toxin production (Zhang et al., 2000) or production of host transforming phages like ɸ13. Antibiotics with a MIC/MTC ratio below 1.0 would be safe in this regard; however, as shown for hydrogen peroxide (Figure 5), the MIC/MTC ratio can be strain or species-specific. Timing of the induction process revealed a highly stochastic process with a maximal toggling rate, leading to dedicated cells with the stable "lytic" phenotype, between 20 and 30 min of exposure to mitomycin C. While F I G U R E 7 Induction of "lysogenic" to "lytic" toggle switching from pCSK9 plasmids in S. aureus by temporal exposure to mitomycin C. "Lysogenic" 8325-4 geh::pCSK9 (CSK48) cells were grown in liquid TSB medium (in duplicates) and induced by addition of 1 µg ml −1 mitomycin C at time = 0 min. At the indicated times, aliquots were harvested and dilutions were plated on TSA plates containing X-gal. The average fraction of colonies (in percent) with the "lytic" phenotype was plotted against the duration (in min) of the exposure to mitomycin C the induction of DNA damage in each cell ultimately defines a particular CI degradation rate, this could easily be changing over time due to the SOS induction of DNA repair enzymes. The stochasticity of the toggle frequency at different time points could originate at many levels, including the initial CI and MOR concentrations, their production levels, and the DNA damage-induced CI degradation rate. While toggle switching has been reported for many bistable switches (Dubnau & Losick, 2006), the ability of an isolated switch to conduct both decision switching and toggle switching has not previously been demonstrated for any phage.

E TH I C S S TATEM ENT
None required.

ACK N OWLED G EM ENTS
We thank associate professor Peter W. Thulstrup from the University of Copenhagen for access to the CD spectropolarimeter. A.K.V.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data generated and analyzed during the current study are available in the Zenodo repository at https://doi.org/10.5281/ zenodo.5515850

A PPEN D I X F I G U R E A 1
The protein sequence of ɸ13 rCI used in this study. The added tag is marked with a black box and consists of the His-tag (HHHHHH) and a TEV cleavage site (ENLYFQG). Cleavage of the tag leaves the G F I G U R E A 2 Sequence alignments of CI from L. lactis phage TP901-1, S. aureus phage ɸ13, and E. coli phage λ (top) and MOR from TP901-1 and ɸ13 (bottom). Identical and similar residues are shown in gray. Known domains are indicated by lines and designated by brief descriptions. The overall similarity between the CI repressors from ɸ13 and TP901-1 is 51% identical residues, while the local domain similarity for the NTD and CTD 1 is 71% and 39%, respectively. The overall similarity between CI from ɸ13 and λ is 19% identical residues, where the N-terminal parts only show similarities in the DNA binding domains. In the C-terminal part, where λ carries its peptidase domain required for autocleavage, the local similarity is 21%, but the catalytic Ser and Lys residues and the G-residue of the autocleavage site are conserved. The overall similarity for the MOR anti-repressors from ɸ13 and TP901-1 is 65% Note: Switch plasmids were replicative (pCSK13) or required integration into the host (pCSK12 in S. aureus, pCSK11 in B. subtilis). pCSK11-13 are mor-deficient versions of pCSK2, pCSK9, and pCSK3, respectively. Recipient bacteria were naïve (S. aureus 8325-4 or B. subtilis 168) or ɸ13 lysogens (S. aureus 8325-4 ɸ13-kana). The number of biological replicates with more than 100 resulting colonies are stated in brackets. Frequencies of "lysogenic" transformant colonies on plates containing X-gal were calculated as the frequency between "lytic" and "lysogenic" in each transformation event and the average is shown here. Standard deviations are based on differences between each event.

F I G U R E
a Indicates that <100 transformants obtained in some or all experiments. The total number of colonies is stated in brackets.