A regulatory cascade controls Staphylococcus aureus pathogenicity island activation

Abstract

Staphylococcal pathogenicity islands (SaPIs) are a family of closely related mobile chromosomal islands that encode and disseminate the superantigen toxins, TSST-1 and SEB. They are regulated by master repressors, which are counteracted by helper phage-encoded proteins, thereby inducing their excision, replication, packaging and intercell transfer. SaPIs are major components of the staphylococcal mobilome, occupying 5 chromosomal att sites, many strains harbouring 2 or more. As regulatory interactions between co-resident SaPIs could have profound effects on the spread of superantigen pathobiology, we initiated the current study to search for such interactions. Using classical genetics, we found that with one exception, their regulatory systems do not cross-react. The exception was SaPI3, which was originally considered defective because it could not be mobilized by any known helper phage. We show here that SaPI3 has an atypical regulatory module and is induced not by a phage but by many other SaPIs, including SaPI2, SaPIbov1 and SaPIn1, each encoding a conserved protein, Sis, which counteracts the SaPI3 repressor, generating an intracellular regulatory cascade: the co-resident SaPI, when conventionally induced by a helper phage, expresses its sis gene which, in turn, induces SaPI3, enabling it to spread. Using bioinformatics analysis, we have identified more than 30 closely related co-ancestral SEB-encoding SaPI3 relatives, occupying the same att site and controlled by a conserved regulatory module, immA-immR-str’. This module is functionally analogous to but unrelated to the typical SaPI regulatory module, stl-str. As SaPIs are phage satellites, SaPI3 and its relatives are SaPI satellites.

Introduction

Horizontal gene transfer (HGT) can in one step transform an opportunistic pathogen into a highly virulent one. In S. aureus, HGT enables relatively benign staphylococcal strains to cause lethal toxic shock, necrotizing pneumonia and necrotizing fasciitis^1,2. Key agents of HGT in staphylococci are the staphylococcal pathogenicity islands (SaPIs)¹, widespread chromosomally integrated ~15 kb elements, with many staphylococcal strains containing two or more^3,4. SaPIs encode and disseminate superantigens including enterotoxin B (SEB) and the toxic shock syndrome toxin (TSST-1), plus antibiotic resistances and other virulence factors^3,4. SaPIs are maintained in a prophage-like state by a master repressor and are induced by helper phages, resulting in the excision, replication and packaging of SaPI DNA in phage-like particles composed of helper phage virion proteins^3–5. SaPIs are controlled by a prophage-like regulatory module consisting of divergently transcribed master repressor and activator genes, stl and str, which correspond to λ c1 and cro. But, unlike most of the classical phage repressors, Stl is not cleaved following activation of the SOS response. Instead it is counteracted by complexing with a helper phage anti-repressor protein^6–9. This couples the SaPI life cycle with that of the helper phage, ensuring that the SaPI is not activated unless the reproductive cycle of a helper phage is in progress. The Stl sequences are rather poorly conserved among the SaPIs and different SaPIs utilize different phage proteins for induction^6–8. The co-existence of 2 or more SaPIs in a single strain raised the question of whether these closely related elements interact, since any such interaction could greatly impact the spread of these rather dangerous genetic elements among staphylococcal strains, possibly with adverse clinical consequences.

Results

Interactions between co-existing SaPIs

We initiated this study to determine whether such interactions occur and, if so, how they influence the induction and transfer of one another, using the well-studied SaPIs: 1, 2, 3, and, bov1, which occupy att sites II (SaPIbov1), IV (SaPI1 and SaPI3), and V (SaPI2)¹⁰, plus their helper phages, 80α, and Φ11. We constructed double-SaPI strains with several differentially marked pairs (these had to be from different att sites because SaPIs using the same att site are mutually exclusive). In cases where a helper phage was known to induce one of a pair but not the other, the objective was to determine whether the non-induced one was impacted by induction of the induced one, or whether the non-induced one impacted induction of the inducible one. The strains were infected with the different helper phages and the lysates scored for transfer frequency of each of the pairs compared with the individual SaPIs alone. The results with 3 of the possible pairs (SaPI1 vs SaPI2, SaPI1 vs SaPIbov1, and SaPI2 vs SaPIbov1) were clear: induction of one member of the pair had no effect on transfer of the other, that is, there was no obvious cross interaction with respect to induction and transfer between SaPIs occupying different att sites (Table S1). When both members of the pair were induced the numbers were roughly equivalent, with the total number of particles being about the same as with either alone. The results with SaPI3, however, were dramatically different. SaPI3 had been considered defective because tests with many different staphylococcal temperate phages had failed to identify a single one that could induce it (Table S2&3). However, when paired with either SaPI2 or SaPIbov1, SaPI3 was always co-induced, suggesting that each of them could cross-induce it (Table 1, Figure 1 a&b, Figure 2a&b). This was confirmed with phage 80α Δdut which cannot induce SaPIbov1⁶ but retains the ability to induce SaPI2. The 80α Δdut was unable to induce SaPI3 via SaPIbov1 (Fig. 1a-c) but was still able to induce SaPI3 via SaPI2 (Figure 2 a&b, Table 1).

Table 1. Interactions of co-resident SaPIs.

	SaPIs
Phage	SaPIbov1	SaPI2	SaPI3	SaPIbov1 / SaPI2	SaPIbov1 / SaPI3	SaPI2 / SaPI3
80α	+	+	-	+ / +	+ / +	+ / +
80α Δdut	-	+	-	- / +	- / -	+ / +
ϕ11	+	-	-	+ / -	+ / +	- / -

”+” = induced; “-“ = not induced. Some entries are shaded to emphasize their importance.

Figure 1. Helper SaPIbov1 promotes satellite SaPI3 induction and transfer.

Lysogenic S. aureus strains carrying phages 80α, 80α Δdut or ϕ11, and the indicated SaPI(s), were MC induced and titres of (a) SaPI3 seb::ermC and (b) SaPIbov1 tst::tetM determined by transduction into RN4220. Error bars indicate ±s.d. (n=4 (80α and 80α Δdut lysogens) or n=3 (Φ11 lysogens) biological replicates). Statistical analysis was performed using One-Way ANOVA on log₁₀ transformed data followed by Sidak’s multiple comparisons test. p-values are as follows: * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, ns not significant. Additional information is provided in the supplementary data file. Diamonds located at the baseline indicate replicates with no observed transductants. (c) Representative Southern blot analyses (n=3) of the strains analysed in panels (a) and (b). Samples were isolated 60, 90 or 120 min after induction with MC, separated on agarose gel and blotted with SaPI3-(top panel) or SaPIbov1- (bottom panel) specific probes. The upper band is ‘bulk’ DNA, including chromosomal, phage, and replicating SaPI; the lower band is SaPI linear monomers (L) released from SaPI-sized phage heads. (d) Representative Southern blot analyses (n=3) of SaPIbov1 and SaPI3 excision and replication following induction of the cloned 80α dut gene. The strains containing SaPIbov1 tst::tetM, SaPI3 seb::ermC or both islands were complemented with plasmid pJP2511 (empty plasmid) or its derivative expressing the 80α Dut protein (pJP2533). Samples were isolated 4 hours after induction with 5 μM CdCl2, which induces dut expression, and Southern blots were performed using SaPI3- and SaPIbov1-specific probes. The upper band is ‘bulk’ DNA. In these experiments, because no helper phage is present, the excised SaPI DNA appears as covalently closed circular molecules (CCC) rather than the linear monomers that are seen following helper-phage-mediated induction and packaging.

Figure 2. Impact of SaPI3 induction on helper SaPIs replication and transfer.

S. aureus lysogenic for 80α containing the indicated SaPIs were MC induced and the SaPI3 seb::ermC (erythromycin resistance marker) (a) or helper SaPI (tetracycline resistance marker) (b) titres were determined. Error bars indicate ±s.d. (n=3 biological replicates). Statistical analysis was performed using One-Way ANOVA on log 10 transformed data followed by Sidak’s multiple comparisons test. * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, ns not significant. Additional information is provided in the supplementary data file. (c) Representative Southern blot analyses (n=3) analysis of mini-lysates from the defined strains taken 2 h after MC induction using probes specific for either SaPI3 (top panel), or the tetM marker of the helper SaPIs SaPIbov1, SaPI2, SaPIn1 or a SaPIbov2-specific probe (bottom panel).

To confirm that the helper phage had no role beyond inducing the helper SaPI, we tested the co-resident SaPIbov1/SaPI3 pair without any phage but with the cloned 80α dut, which expression activates SaPIbov1⁶, and found that both SaPIs were induced (Fig. 1d).

Identification of the SaPI3 inducer

To identify the SaPIbov1 gene(s) responsible for SaPI3 induction we used an existing set of SaPIbov1 deletion mutants¹¹. The mutants were individually introduced into RN4220 containing SaPI3 seb::ermC. The strains were infected with 80α, and SaPI3 transfer frequency was measured. Deletion of SaPIbov1 ORF16 completely eliminated high-frequency SaPI3 transfer (Table S4) suggesting that SaPIbov1 gp16 was the inducer of SaPI3. To confirm this, we cloned ORF16 under the control of the cadmium-inducible promoter (Pcad), tested it for SaPI3 induction in a non-lysogenic strain containing only SaPI3, and found that expression of ORF16 was sufficient to induce SaPI3 excision and replication (Fig. S3). This gene was accordingly named sis (SaPI inducer of SaPIs).

Generality of SaPI3 induction

As SaPIbov1 Sis is conserved among the SaPIs (Table S5 and Fig. S4), we next questioned whether any of the other SaPI-encoded Sis homologs could serve as SaPI3 inducers. We cloned the sis homologs from SaPI1, SaPI2, SaPI3, SaPI5, SaPIn1, SaPIbov1 and SaPIbov2 under the Pcad promoter in expression vector pCN51, expressed the different variants in a SaPI3-positive strain, and observed that all but two of them (SaPIs 5 and bov2 sis, which are considerably divergent) induced SaPI3 (Fig. S3; Fig. S4). Like SaPI2 and SaPIbov1, SaPIn1 could also induce SaPI3, whereas SaPIbov2 could not confirming that the behaviour of cloned sis mirrors that of the intact SaPIs (Fig. 2). SaPIs 1 & 5 could not be tested because they use the same attB site as SaPI3.

SaPI3 regulatory system

Like most other SaPIs, SaPI1 has a conserved, two gene regulatory core and an intervening control region containing promoters and operators. In SaPI3 the regulatory core is replaced by three genes (Fig. S1), including homologues of immA and immR, which regulate the excision and transfer of integrative and conjugative (ICE) elements such as ICEBs1 from Bacillus subtilis, and are present in many staphylococcal phages¹² (Table S6). The third gene corresponds to the SaPI str gene but is unrelated to any previously identified SaPI str gene and is here designated str’. We hypothesised that this module represents the SaPI3 control system and is critical for the remarkable regulatory behaviour of SaPI3 as a satellite. ImmR would thus be the SaPI3 repressor, analogous to Stl, and Str’ the lambda Cro analogue, corresponding to Str. Other SaPIs lack an ImmA analogue. In B. subtilis ImmA has protease activity that cleaves ImmR and in B. subtilis either the SOS response or the RapI-PhrI sensory system activate ICEBs1 transfer via ImmA degradation of ImmR. Since the immA-immR complex is SOS-inducible in its ICE context¹², we tested for SOS-induction of SaPI3 even though SOS inducibility would be counter-selective for a SaPI. As expected, SaPI3 was not induced to excise and replicate by mitomycin C (MC) (Fig. 1c and 2c). Importantly, SaPI3 is the prototype of an extensive lineage containing over 30 closely related SaPIs in which this exclusive regulatory module is highly conserved (see Fig. S5). These all occupy the same attB site, and most encode SEB.

ImmR is the SaPI3 master repressor

To determine whether ImmR is the master repressor for SaPI3, we attempted to isolate an immR deletion but were unsuccessful, perhaps because loss of ImmR results in toxic uncontrolled SaPI3 replication. To bypass this problem, we constructed a set of test plasmids containing different components of the immA/immR/str’ module. One of these, pJP2524, contained immR under control of its native promoter plus the intergenic region between immR and str’, with a β-lactamase reporter gene (blaZ) driven by Pstr’. This construct expressed blaZ at a barely detectable level, compared to a second plasmid, pJP2523, containing only the intergenic region with blaZ driven by Pstr’, which expressed blaZ at a high level (Fig. S6a and b). This was blocked by the expression of immR cloned to a second plasmid under Pcad control (Fig. S6c).

Sis interacts with ImmR to relieve SaPI3 repression

The fact that ImmR works like the classical Stl repressors suggests that Sis binds directly to ImmR. Since Sis homologs were found to be either insoluble or toxic in E. coli, we were unable to test for this by the E. coli 2-hybrid assay. Instead, we used super-resolution microscopy with C-terminally-tagged fluorescent derivatives, SaPIbov1 Sis with mCherry and SaPI3 ImmR with GFP, and confirmed that both were functional (Fig. S7). The resulting plasmids were introduced into RN4220, the Sis-mCherry/ImmR-GFP proteins expressed, and the samples analysed (Fig. 3). When both mCherry-tagged Sis and GFP-tagged ImmR were expressed within the same cell, the fluorescent signals were clearly co-localized (Fig. 3a). These data are consistent with a direct interaction between Sis and ImmR, as would be expected for relief of ImmR repression of the SaPI3 promoter. As controls the pairs mCherry/ImmR-GFP or Sis-mCherry/GFP were also expressed. In the absence of Sis, the fluorescent signals for mCherry and ImmR-GFP showed diffuse distribution throughout the staphylococcal cell (Fig. 3b). By contrast, the fluorescent signal of the mCherry-tagged Sis protein was localised within the staphylococcal cell while GFP alone remained diffuse (Fig. 3c), showing that these did not interact.

Figure 3. Sis and ImmR colocalise in vivo.

S. aureus RN4220 expressing either (a) SaPIbov1 Sis-mCherry and SaPI3 ImmR-GFP, (b) mCherry and SaPI3 ImmR-GFP or (c) SaPIbov1 Sis-mCherry and GFP were induced with 5 μM CdCl₂ for 16 h and samples prepared and subjected to super-resolution microscopy. Maximum projections (top panels, scale bars 5 μm) and representative 3D-reconstructions (bottom panels, scale bars 0.2 μm) for each sample are shown. Depicted colours are false coloured. Clear: cell wall stained with wheat germ agglutinin Alexa Fluor 467 (WGA); purple: mCherry; yellow: GFP. Representative images from 3 independent experiments are shown.

To test for the generality of repression relief by Sis, we generated a set of strains containing two plasmids: the aforementioned plasmid pJP2524, which carries a β-lactamase reporter gene whose expression is repressed by ImmR, and each of the pCN51 derivative plasmids expressing the Cd-inducible sis gene from a different island (see scheme in Fig. 4a). The empty pCN51 plasmid was used as a control. Expression of the SaPI1, SaPI2, SaP3, SaPIn1 and SaPIbov1 sis homologs, but not the SaPIbov2 or SaPI5 variants, strongly increased β-lactamase expression from the ImmR-repressed str’ promoter (Figure 4b), consistent with an interaction between Sis and ImmR. To demonstrate the SaPI3 specificity of this interaction, we tested for induction by the cloned Sis proteins of expression of the cloned str promoters of the classical SaPIs, SaPI1, SaPIbov1 and SaPI2, and found that none of the Sis homologs was able to induce expression from these promoters (Fig. S8). Taken together, these results show that the helper SaPIs induce the SaPI3 cycle by producing a protein, Sis, that alleviates the ImmR-mediated repression, a mechanism that mirrors the derepression of the classical SaPIs by phage-encoded anti-repressors^6,8.

Figure 4. The helper SaPI Sis inducer alleviates SaPI3 ImmR repression.

(a) Schematic representation of the reporter and expression plasmids used. The β-lactamase reporter gene (blaZ) is under the control of SaPI3 ImmR, in presence or absence of ImmA. Expression of different cloned sis genes is under the control of a cadmium-inducible promoter (Pcad). (b) Strains containing either pJP2524 or pJP2525, as well as any of the different pCN51-derivative plasmids expressing the different Sis homologs, were assayed for β-lactamase activity 5 h after induction with 1 μM CdCl₂. (c) Strains containing pJP2526 and any of the different pCN51-derivative plasmids expressing the different Sis homologs were assayed for β-lactamase activity 5 h after induction with 1 μM CdCl₂. Error bars indicate ±s.d. (n=3 biological replicates). Statistical analysis was performed using One-Way ANOVA followed by Holm-Sidak’s multiple comparisons test. p-values are as follows: * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, ns not significant. Additional information is provided in the supplementary data file.

ImmA is required for SaPI3 induction

As Sis is sufficient to induce SaPI3, what is the role of SaPI3 ImmA in this process, especially since ImmA is required for ICE induction in B. subtilis ¹²? To answer this question, we deleted SaPI3 immA and tested for activation of the mutant SaPI3 by 80α-induced SaPIbov1. Although SaPI3 ΔimmA was de-repressed, the de-repression was only partial, suggesting that ImmA was required for the full-scale activation of SaPI3 (Fig. 5a).

Figure 5. ImmA is required for full activation of SaPI3.

(a) 80α lysogens carrying SaPIbov1 and the indicated SaPI3 islands were MC induced and the transfer of SaPI3 assessed in RN4220. (b) The SaPI3 immR or stir’ genes were cloned into pUT18c (pJP2531 or pJP2532, respectively) and the immA gene was cloned into pKNT25 (pJP2530). The pUT18c- and pKNT25-derivative plasmids were co-transformed into E. coli strain BTH101. Serial dilutions of an overnight culture were plated onto LB supplemented with kanamycin, ampicillin and 100 μM IPTG and 20 μg ml^-1 X-gal. BTH101 transformed with pUT18c-zip and pKNT25-zip or pUT18c and pKNT25 served as positive or negative controls for protein-protein interactions, respectively. (c) RN4220 80α lysogens carrying the SaPI3 seb::tetM WT or the ΔimmA mutant were transformed with either an empty control plasmid (pCN51), a plasmid expressing the wild-type ImmA (pJP2520) or a plasmid expressing an ImmA version mutated in the protein’s active site (pJP2521, ImmA E51A/E52A). Cultures were grown to early exponential phase, MC treated and the expression of ImmA was induced with 1 μM CdCl2. To analyze SaPI3 transfer, the lysates were transduced in RN4220. Error bars indicate ±s.d. (n=3 biological replicates). Statistical analysis was performed using log₁₀ transformed data and an unpaired, two-sided Student’s t-test (a) or One-Way ANOVA followed by Sidak’s multiple comparisons test (c). p-values are as follows: * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, ns not significant. Additional information is provided in the supplementary data file.

To gain insight into the role of ImmA in SaPI3 induction, we first determined whether immA expression increases after sis induction - which would be predicted since immA and immR are almost certainly co-transcribed. For this purpose, we used plasmid pJP2526, which carries a β-lactamase reporter gene fused to the immR/immA segment (see scheme in Fig. 4a). The several sis-expressing plasmids described above were introduced into cells carrying this plasmid and the expression of the β-lactamase reporter tested upon CdCl₂ induction. Expression of the sis homologs (from SaPIs 1, 2, 3, SaPIbov1 and SaPIn1) that were able to induce SaPI3 also increased the expression of immA by relieving ImmR-mediated repression (Fig. 4c).

Next we analysed how the presence of ImmA affects the induction by Sis of str’ expression. We repeated the aforementioned experiment with pJP2524, in which the expression of the blaZ reporter, blocked by ImmR, was measured in presence of the different inducers (Fig. 4a-b), but now using a plasmid construct which also contains the immA gene adjacent to the immR gene, mirroring the structure present in SaPI3 (plasmid pJP2525, see scheme in Fig. 4a). In support of the previous results, str’ expression was significantly greater in the presence of ImmA than in its absence (Fig. 4b), confirming the positive effect of this protein on str’ activation.

ImmA is an inducer of SaPI3

The above result raised several questions: Does ImmA interact directly with ImmR? Is the presence of Sis required for this interaction? Does ImmA induce SaPI3 directly? Does ImmA degrade ImmR as is the case with the ICEBs1 element? A two-hybrid assay demonstrated that ImmR and ImmA interact directly, in the absence of Sis (Fig. 5b). Most significantly, a test of immA cloned under the Pcad promoter (pJP2520) demonstrated that independently cloned immA can induce SaPI3 in the absence of any Sis-producing helper SaPI. This enables 80α to promote the high-frequency transfer of SaPI3 (Fig. 5c). Additionally, a test of a protease-deficient mutant of ImmA (E52A/E53A¹²), showed that the protease activity of ImmA was important but not completely required for its SaPI3 inducing activity (Fig. 5c). It is concluded from these results that ImmA is an alternative de-repressor of SaPI3, but only when its production is initiated by Sis-mediated de-repression – which sets up a positive feedback loop for SaPI3 induction (Fig. 4c). As no such feedback loop appears to be involved in the induction of the classic SaPIs, this makes the SaPI3 mode of regulation unique, and suggests that it may be selectively advantageous.

In summary, our results indicate that ImmR represses both the rightward (str’) and leftward (immR) promoters in the intergenic region, and thereby also represses immA. Sis binds to ImmR, relieving its repressor activity on both promoters. The resulting activation of the str’ promoter initiates the SaPI3 replication cycle. However, Sis-induced relief of the ImmR-specific repression is not sufficient for full activation, and the de-repression is completed by binding of ImmA to ImmR and by proteolytic cleavage of ImmR by ImmA (Fig. 6).

Figure 6. Model of SaPI3 activation.

(a) SaPI3 is maintained in the chromosome by its ImmR repressor, which represses expression from both PimmR and Pstr’ promoters. (b) Helper SaPI Sis proteins (1) bind to ImmR and sequester it from its DNA target (2) allowing initial expression of SaPI3 replication genes from the str’ promoter as well as of ImmA (3) from the immR promoter. (c) ImmA can either bind to ImmR and remove it from its DNA target (4a) or can bind to the complex of ImmR and Sis (4b) leading to ImmR degradation (5a&b). Although this has not been experimentally tested, once induction of SaPI3 has started, the island can provide its own Sis (6) to sequester any ImmR produced and ensure replication.

SaPI3-encoded Sis enhances SaPI3 induction

SaPI3 encodes its own Sis homolog (ORF14), which is able to induce SaPI3 activation better than any of the other Sis homologs tested (Fig. 4 and S3). To test whether SaPI3 Sis was involved as a positive feedback mechanism reenforcing SaPI3 activation once replication of SaPI3 was triggered by an exogenous Sis homolog, we constructed a SaPI3 Δsis mutant lysogenic for 80α and expressed either the SaPIbov1 or SaPIn1 Sis homologs from a cadmium-inducible plasmid representing a strong and weak inducer of SaPI3, respectively. These strains were MC induced and the SaPI3 transfer analysed. SaPI3 titres were significantly reduced in the SaPI3 Δsis mutant strain irrespective of the exogenous Sis kickstarting the system (Fig. S9) confirming the existence of a positive feedback loop for SaPI3 activation.

Role of Sis in the SaPI cycle

The fact that all the analysed SaPIs encode a Sis homolog suggests these proteins may have an important role in the SaPI cycle. In a previous study we demonstrated that deletion of sis did not affect SaPIbov1 replication¹¹. To extend our studies to other islands, we deleted sis in SaPI2 and SaPIn1, and tested the transfer of these mutant islands after their induction by helper phage 80α. Deletion of the SaPI2 sis gene did not impact either SaPI2 transfer or replication (Fig. S10a and b). However, while SaPIn1 replication was unaffected, deletion of sis severely reduced SaPIn1 transfer (Fig. S10), suggesting that Sis may have an important role in the SaPI life cycle. Reconstitution of sis in the SaPIn1 Δsis mutant fully restored WT levels of SaPIn1 transfer confirming that the observed reduction in SaPIn1 transfer titres was the result of the SaPIn1 sis deletion (Figure S10).

SaPI3 interference with helper SaPI and helper phage

By analogy with the interference by SaPIs with helper phage reproduction^13–16, SaPI3 significantly interfered with the transfer of the helper SaPI2 and SaPIbov1 islands (Fig. 1b and 2b). This interference, however, did not occur with SaPIn1 (Fig. 2b). Interestingly, Southern blot analyses confirmed that this interference occurred at the level of helper SaPI DNA replication, as the presence of induced SaPI3 substantially reduced SaPIbov1 and SaPI2 DNA replication (Fig. 1c and 2c). The precise mechanism by which SaPI3 interferes with its helper SaPI is now under study, but it clearly involves a different mechanism than those used by the SaPIs or PICIs to block phage reproduction, all of which targeted phage packaging^13–16. Additionally, as SaPI3 contains the same phage interference genes as other SaPIs, it would be predicted to interfere with helper phage as well as helper SaPI reproduction. Accordingly, we compared the titres of the 80α and 80α Δdut phages after infection of a SaPI-negative strain or the same strain carrying either SaPI3, SaPIbov1 or both. In agreement with previous results¹⁷, SaPIbov1 induction by 80α slightly interfered with phage reproduction (Fig. S11a and b). As expected, since 80α does not induce SaPI3, the 80α titre was unaffected by the presence of this island alone. Remarkably, the presence of SaPIbov1 and SaPI3 in the same strain severely interfered with 80α reproduction, suggesting a cooperative anti-phage activity. Interestingly, whereas induction of the satellite SaPI was detrimental for the helper one, the helper-satellite interaction was beneficial for the host strain, since it increased the number of cells that survived the phage infection (Fig. S11b).

Discussion

SaPI3 had no known helper phage when this study began. Our discovery that the Sis proteins of many (but not all) SaPIs function as an anti-repressor for SaPI3 revealed an intriguing paradigm: a three-part cascade begins with a helper phage that induces a helper SaPI, which expresses Sis, which activates the satellite SaPI3. The regulatory module of SaPI3, consisting of 3 genes, includes homologs of immA and immR which are common in phages and other MGEs and together regulate an adjacent transcriptional activator, str’, corresponding to the standard SaPI str, to which it is unrelated. This module is present in a large number of similar or identical SaPIs (Fig. S5), all located in the same chromosomal site, often encoding enterotoxin B and having virtually identical genome organisation. This suggested that the recombinational event occurred once and that these SaPIs comprise a lineage consisting of evolutionary progeny of the initial recombinant. We suggest that this remarkable paradigm is important for SaPI biology as the SaPI3-like and helper SaPIs interact with one another and with helper phages in interesting ways that could contribute importantly to the transfer and prevalence of these SaPIs.

The genomes of phages, SaPIs and other mobile elements are mosaic owing to remarkable recombinational promiscuity; against this background are a few one-of-a-kind recombination events, that have occurred by unknown mechanism(s) and have had large effects on the elements involved. SaPIs 1 & 3 have acquired genes twice, SaPIbov5 once. In SaPI1 (carrying tst) and SaPI3 (carrying seb), tst and seb were likely acquired by recombination and occupy precisely the same site but in opposite orientations. Also, in the otherwise largely identical genomes of SaPIs 1 & 3, the regulatory core of SaPI1 was precisely replaced, in SaPI3, with a novel regulatory core with the same transcriptional organization and overall function, but dramatically different mechanics, eliminating induction by helper phages while enabling induction by Sis of co-resident SaPIs, as described in this report. SaPIbov5 is the result of recombinational replacement of most of the operon I¹⁸ of a classic SaPI with a segment of cos phage DNA, converting it from a pac element to a cos element. As this replacement event interrupted its terS gene, SaPIbov5 is totally dependent on a helper phage for the packaging of its DNA¹⁹, as well as for its induction.

The evolution of the SaPI3 system, which resulted in the coexistence of one MGE with a second one that encodes a moonlighting protein that de-represses the first, must be viewed as entirely parallel with that of the initially described SaPI-helper phage combination. In other words, once evolved from a prophage, the SaPI3 lineage evolved away from SOS induction because SOS induction would be highly counter-selective for such a phage-like element in the absence of a packaging and transfer system. A likely evolutionary pathway would then be the occurrence of repressor mutations that enabled binding by a protein encoded by a co-resident SaPI rather than a prophage.

The immR/immA regulatory system used by SaPI3 differs from classic SaPI regulation not only in responding to SaPI-encoded anti-repressors, but also in the possession of a double autoinduction circuit. This autoinduction starts with the initial Sis-induced trans-derepression of the leftward immR promoter, resulting in the expression of immA. ImmA binds to ImmR and degrades it, which further up-regulates the immR promoter. The combined action of ImmA and helper-produced Sis causes activation of the rightward-facing str’ promoter. Among the genes that are activated is ORF14, encoding the SaPI3 Sis – the most active of the several Sis proteins tested (see Fig. 4 and S3). Absence of SaPI3 sis reduced transfer of SaPI3 irrespective of the Sis homolog used for induction (Fig. S9). Thus, SaPI3 sis is required to ensure full induction of the SaPI3 replication cycle. Once the helper-encoded Sis gets the system started, the autoinduction circuit does the rest.

Does SaPI-encoded Sis act on other MGEs that are regulated by the immA-immR system? If so, it is feasible that cross-regulating MGEs might have important biological effects on HGT, acquisition of virulence and antibiotic resistance.

Online Methods

Bacterial strains and culture conditions

The bacterial strains used in this study are detailed in Table S7. S. aureus strains were grown in Tryptic soy broth (TSB) or on Tryptic soy agar plates and E. coli strains were grown in Luria-Bertani broth (LB) or on LB agar plates. Antibiotic selection was used where appropriate (tetracycline 3 μg ml^-1, erythromycin 10 or 2.5 μg ml^-1 (as indicated), chloramphenicol 10 μg ml^-1, ampicillin 100 μg ml^-1, kanamycin 30 μg ml^-1).

DNA manipulations

General DNA manipulations were performed using standard procedures. Plasmid constructs used in this study (Table S8) were generated by cloning PCR products (Kapa Hifi Polymerase, Roche) obtained with oligonucleotide primers listed in Table S8. Detection probes for SaPI DNA in Southern blots were generated by PCR using a non-proofreading polymerase (DreamTaq polymerase, ThermoFisher) using oligonucleotides specified in Table S9. Probe labelling and DNA hybridization were performed following the protocol provided with the PCR-DIG DNA-labelling and chemiluminescent detection kit (Roche).

Southern blotting

Strains containing the defined phage, SaPI(s) and/or plasmids were grown to early exponential phase (OD₅₄₀~0.15) in 10 ml of TSB supplemented with antibiotics where plasmids were present. Phages were induced with mitomycin C (2 μg ml^-1) and, where pCN51 expression plasmid derivatives were present, 1-5 μM CdCl₂ was added to induce expression as indicated. One ml samples were taken at the defined timepoints, pelleted by centrifugation (16000 × g, 2 min) and shock frozen on dry ice. The samples were re-suspended in 50 μl lysis buffer (47.5 μl TES-Sucrose (10 mM Tris-Cl, 100 mM NaCl, 1 mM EDTA, 20% (w/v) sucrose) and 2.5 μl lysostaphin [12.5 μg ml^-1]) and incubated at 37°C for 1 h. Following the incubation, 55 μl of SDS 2% proteinase K buffer (47.25 μl H2O, 5.25 μl SDS 20%, 2.5 μl proteinase K [20 mg ml^-1]) was added before incubation at 55°C for 30 min. Samples were vortexed for 1 h with 11 μl of 10x loading dye followed by three cycles of 5 min incubations in dry ice/ethanol and at 65°C in a water bath. Samples were run on 0.7% agarose gel at 25-30 V overnight. DNA was transferred by capillary action to a positively charged nylon membrane (Roche), processed as per the manufacturer’s instructions, and exposed using a DIG-labelled probe (see DNA methods) and anti-DIG antibody (1:10000 (v/v), Roche, product 11093274910) before washing and visualisation.

Generation of clean deletion mutants and reconstitution of SaPIn1 Δsis mutant

The SaPI3 immA mutant was constructed using allelic replacement by cloning flanking regions up- and downstream of the respective gene into pMAD using oligonucleotide primers described in Table S9. The plasmids were then transformed into strains carrying the SaPI of interest and integration of the plasmid was selected by growth at the restrictive temperature (42°C) on TSA plates supplemented with 80 μg ml^-1 X-gal and 2.5 μg ml^-1 erythromycin and 3 μg ml^-1 tetracycline. Single crossover events were isolated (light blue colonies) and grown overnight under replication-permissive conditions (TSB, 30°C, 80 rpm, 3 μg ml^-1 tetracycline) to facilitated excision and loss of the integrated plasmid. Serial dilutions of the cultures were plated on TSA plates supplemented with 3 μg ml^-1 tetracycline and 80 μg ml^-1 X-gal and correct deletion mutants identified by PCR followed by sequencing using oligonucleotides annealing outside of the recombination region.

SaPI transduction and generation of SaPI-containing strains

SaPIs (except SaPI3) were transduced into the recipient strains by inducing donor SaPI-containing derivatives of RN4220 lysogenic for 80α ΔterS, which specifically induces and packages resident SaPIs. The SaPIs were then transduced into the desired recipient strain as described below. For SaPI titre determination or SaPI transduction, S. aureus RN4220 or other recipient strains were grown overnight at 37°C and 120 rpm. The culture OD was adjusted to OD₅₄₀~1.4 with TSB and supplemented with 4.4 mM CaCl₂. 100 μl of the appropriate lysate dilution were added to 1 ml of this cell suspension and incubated for 20 min at 37°C. Three ml of transduction top agar (TTA, 30 g l^-1 TSB, Oxoid, 7.5 g l^-1 agar, Formedium) were added to the transduction and the mix poured onto a TSA plate containing the appropriate antibiotic. Plates were incubated for 16-24 at 37°C prior to determination of transducing units (TU ml^-1).

Phage infection

Strains containing the desired donor SaPI were grown in TSB to early exponential phase (OD₅₄₀~0.15) at 37°C and 120 rpm. 10 ml of this culture were pelleted by centrifugation (3300 x g, 10 min) and the pellet was resuspended in 5 ml of fresh TSB and 5 ml of phage buffer (1 mM MgSO₄, 4 mM CaCl₂, 50 mM Tris-Cl, 100 mM NaCl, pH=8). Phages were added to the culture at a multiplicity of infection of 10:1 and infected cultures incubated for 4 h at 30°C and 80 rpm followed by overnight incubation at room temperature. Cleared lysates were filtered using a 0.2 μm syringe filter (Satorius).

Phage induction and titration

S. aureus strains lysogenic for helper phages and containing the required SaPI(s) were grown to early exponential phase (OD₅₄₀~0.15) at 37°C and 120 rpm. Cultures were then induced by the addition of mitomycin C (2 μg ml^-1) and incubated for 4-5 h at 30°C followed by overnight incubation at room temperature before filtering with a 0.2 μm syringe filter (Sartorius). To determine the phage titres, RN4220 cultures were grown to OD₅₄₀~0.35 and 100 μl of this culture were mixed with 3 ml of phage top agar (PTA, 20 g l^-1 Nutrient Broth No. 2, Oxoid, plus 3.5 g l^-1 agar, Formedium supplemented with 10 mM CaCl₂) and overlaid onto phage base agar plates (20 g l^-1 Nutrient Broth No. 2, Oxoid, plus 7 g l^-1 agar, Formedium supplemented with 10 mM CaCl₂). Phage lysates and dilutions in phage buffer (1 mM MgSO₄, 4 mM CaCl₂, 50 mM Tris-Cl, 100 mM NaCl, pH=8) were spotted in triplicates of 10 μl each onto lawns of the specified strains, dried and incubated overnight prior to plaque forming unit (PFU ml^-1) determination.

Two-hybrid assay

The two-hybrid assay for protein-protein interaction was performed as described previously¹⁶ using two compatible plasmids: pUT18c expressing T18 fusion with either SaPI3 ImmR or Str, and pKNT25 expressing the T25 fusion with the SaPI3 ImmA. Both plasmids were co-transformed into E. coli BTH101 for the Bacterial Adenylate Cyclase Two Hybrid (BACTH) system and plated on LB supplemented with ampicillin (100 μg ml^-1), kanamycin (30 μg ml^-1), X-gal (20 μg ml^-1) and IPTG (100 μM). After incubation at 30°C for 48 h (early reaction) to 72 h (late reaction), the protein-protein interaction was detected by a colour change. Blue colonies represent an interaction between the two clones, while white/yellow colonies are negative for any interaction.

Promoter activity assay

For the β-lactamase assays, cultures were grown in 5 ml TSB supplemented with the appropriate antibiotics at 37°C and 120 rpm to early exponential phase (OD₅₄₀~0.15) and 200 μl of culture were added directly to 800 μl of potassium phosphate buffer (50 mM, pH 5.9, supplemented with 10 mM sodium azide) and frozen on dry ice. Where expression of Sis or ImmR was required, cultures were then induced by the addition of 1 μM CdCl₂ and incubated for 5 h at 30°C and 80 rpm followed by collection of 100 μl of culture in 900 μl of potassium phosphate buffer. β-lactamase assays, using nitrocefin as substrate, were performed as described^6,8: 50 μl of the collected sample were mixed with 50 μl of nitrocefin stock solution (192 μM made in 50 mM potassium phosphate buffer, pH 5.9), and immediately reading the absorbance at 490 nm using an ELx808 microplate reader (BioTek) for 30 min. Promoter activity was calculated as Promoter activity = (dA₄₉₀/dt(h))/(OD₅₄₀ x d x V), where OD₅₄₀ is the absorbance of the sample at OD₅₄₀ at collection, d is the dilution factor, and V is the sample volume.

Correlative light-electron microscopy (CLEM)

PCR fragments of SaPIbov1 sis, mCherry, SaPI3 immR and gfp were amplified by PCR using oligonucleotides detailed in Table S8. mCherry was fused to the C-terminus of SaPIbov1 Sis or SaPI3 ImmR and GFP was fused to the C-terminus of SaPI3 ImmR as translational fusion products. The SaPIbov1 sis-mCherry fusion or mCherry alone were cloned into the SalI and KpnI sites of pCN51 generating pAH0537 and pAH0540. Both fusion products as well as a C-terminally mCherry tagged version of immR were also cloned into the SalI and KpnI sites of pAH0556 generating pAH0558, pAH0561 and pAH0560, respectively. The SaPI3 immR-gfp fusion or gfp alone were cloned into the KpnI/AscI sites pAH0537 and pAH0540, respectively, generating plasmids pAH0670, pAH0671 and pAH0679. All plasmids were transformed into RN4220 and transformants were grown to exponential phase at 37°C, 120 rpm prior to induction with 5 μM CdCl₂ for 16 h at 30°C, 80 rpm. 500 μl of each culture was pelleted for 2 min at 16000 x g and washed once with an equal amount of phosphate buffered saline (PBS). The pellet was resuspended in 500 μl PBS supplemented with 4 % (v/v) paraformaldehyde and incubated for 20 min on ice. Cells were collected by centrifugation for 2 min at 16000 x g and the pellet and washed once with and equal volume of PBS. The bacterial cell wall was stained using 2 μg ml^-1 wheat germ agglutinin Alexa Fluor™ 647 Conjugate (Invitrogen) for 10 min, washed twice with 500 μl PBS and resuspended in 100 μl of PBS. 10 μl of cell suspension was spotted onto a microscopy slide and dried on a heat block for 5 min at 50 °C. 4 μl of VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories, Burlingame, California, USA) was added to the cover slip and the cover slip placed onto the dried sample. The cover slip was sealed using transparent nail varnish. Super-resolution Structured Illumination Microscopy (3D-SIM) was performed using a Zeiss Elyra PS.1 super-resolution microscope (Carl Zeiss, Germany). A plan-Apochromat 63x/1.4 oil lens was used, and z-stacks of a total of 2 μm were acquired in three rotations using ZenBlack Edition imaging software (Carl Zeiss, Germany). All images obtained were analysed and processed in Fiji²⁰, where maximum intensity projections were obtained. For 3D model rendering, the z-stacks were processed using Imaris software (Bitplane, Oxford Instrument).

Statistical analyses

Statistical analysis was performed as indicated in the figure legend. In general, phage and SaPI titres were log₁₀-transformed and analysed by either One-Way ANOVA followed by Sidak’s multiple comparison test or using a Student’s unpaired two-tailed t-test as appropriate for the relevant comparison. Promoter activity data were analysed on raw activity data by either One-Way ANOVA followed by Sidak’s multiple comparison test or using a Student’s unpaired two-tailed t-test as appropriate for the relevant comparison. All analysis was done using GraphPad Prism 6 software. The p-values represented in each figure are shown as follows: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns not significant.

Bioinformatic analyses

SaPI sequences were aligned using BLASTp and comparisons were graphed using Easyfig 2.2.5²¹. Annotation and colour-formatting were performed using Adobe Illustrator CS4. Comparison of protein sequence conservation was performed using the PRALINE²² web server (http://www.ibi.vu.nl/programs/pralinewww/help.php). For pairwise comparison of Sis homologs, PRALINE alignments were exported from the web server and analysed in CLC Genomics Workbench (version 7.5.2).

Data availability

No original sequence data was generated in this study. Accession codes for nucleotide sequences harbouring the studied sis homologs are as follows: SaPI1 (U93688), SaPI2 (EF010993), SaPI3 (NC_002951), SaPI4 (NC_002951), SaPI5 (NC_007793), SaPIn1 (NC_002745), SaPIbov1 (NC_007622), SaPIbov2 (AY220730), SaPIbov4 (HM211303), SaPIbov5 (HM228919).