Pathogenicity Island-Directed Transfer of Unlinked Chromosomal Virulence Genes

Summary

In recent decades, the notorious pathogen Staphylococcus aureus has become progressively more contagious, more virulent and more resistant to antibiotics. This implies a rather dynamic evolutionary capability, representing a remarkable level of genomic plasticity, most probably maintained by horizontal gene transfer. Here we report that the staphylococcal pathogenicity islands have a dual role in gene transfer: they not only mediate their own transfer, but they can independently direct the transfer of unlinked chromosomal segments containing virulence genes. While transfer of the island itself requires specific helper phages, transfer of unlinked chromosomal segments does not, so that potentially any pac-type phage will serve. These results reveal that SaPIs can increase the horizontal exchange of accessory genes associated with disease, and may shape pathogen genomes beyond the confines of their attachment sites.

Introduction

Horizontal gene transfer can, in a single step, transform a benign bacterium into a virulent pathogen. This is especially true for the notorious nosocomial and community pathogen, Staphylococcus aureus, for which horizontal gene transfer enables relatively benign strains to cause lethal toxic shock and necrotizing fasciitis (Lindsay et al., 1998; Narita et al., 2001). For most bacterial pathogens, interactions with the host require a complex set of virulence determinants that are often carried by large (up to 200 kb) pathogenicity islands (Hacker and Kaper, 2000). By comparison, the superantigen-carrying Staphylococcal pathogenicity islands (SaPIs) are small (generally 15–18 kb) chromosomally integrated elements that carry few of the genes involved in virulence, but are extremely widespread in S. aureus, with most staphylococcal genomes containing one or more of these elements (Novick et al., 2010). Normally maintained in the host chromosome, SaPIs are induced to excise and replicate by specific helper phages, following the formation of a complex between the SaPI master repressor and phage-coded antirepressor proteins (Tormo-Mas et al., 2010). Though they are highly mobile, the SaPIs lack the machinery for horizontal exchange and exploit specific helper phages for their transfer (Lindsay et al., 1998; Ruzin et al., 2001).

Genome packaging by pac-type phages is initiated with the recognition of a phage-specific packaging site (pac) by the phage small terminase (TerS_Φ), which forms hetero-oligomers with the phage large terminase (TerL) to process viral DNA into procapsids by the headful mechanism (Black, 1989; Casjens, 2011; Rao and Feiss, 2008). To exploit this process, the SaPIs encode their own small terminases (TerS_SP) (Ram et al., 2012; Ubeda et al., 2009), which are highly conserved among the SaPIs, but distantly related to the helper phage terminases, and have highly conserved and unique C-terminal extensions that distinguish them from all known phage terminases (Figure S1A). Yet, despite these differences, SaPI terminases complex with the phage TerL and enable SaPI DNA to be efficiently packaged into phage-derived procapsids and transferred at high frequencies, not only to various S. aureus strains, but also to Listeria monocytogenes (Chen and Novick, 2009).

In S. aureus, most horizontal gene transfer is mediated by temperate phages, which transfer host genomic segments, plasmids and transposons by classical generalized transduction. When phage DNA mispackaging occurs, bacterial host DNA is packaged into procapsids in place of the viral genome, forming generalized transducing particles that can transfer any gene from one bacterium to another (Lennox, 1955; Zinder and Lederberg, 1952). Here we report an unrecognized attribute of the S. aureus pathogenicity islands; namely their ability to contribute to the enhancement of their host’s pathogenicity indepently of their own direct role, by directing the transfer of unlinked genes that contribute to host virulence. We find that homologs of the SaPI pac site are scattered throughout the host chromosome, and are recognized by the SaPI TerS_SP, leading to the encapsidation and high frequency transfer of chromosomal segments downstream of these sites. Moreover, the expression of terS_SP can be induced independently of the SaPI life cycle, in which case specific helper phages are not required so that potentially any pac phage can be used. Since SaPIs are very common in S. aureus, and since they encode terminases with different sequence specificities, they are thus capable as a genre, of promoting the exchange of alleles and the acquisition of a wide variety of genes, many of which are important for pathogen adaptation.

Results

SaPI Mediates Gene Transfer Through Helper Phages

While phage TerS_ϕ mispackaging is infrequent, the sheer scale of phage propagation exceeds the low error rates and drives generalized transduction to appreciable levels. Considering that SaPIs approximate phage-like scales of propagation, we reasoned that if SaPI terminases mispackage DNA at phage-like rates, SaPI-mediated generalized transduction would be detectable.

Previous studies showed that Mitomycin C induction of a strain containing a SaPI and a helper prophage with a terS_ϕ deletion (∆terS_ϕ) results in a lysate containing only SaPI particles (Ubeda et al., 2009), which enables testing for island-mediated generalized transduction (IMT) in the absence of phage-mediated generalized transduction (PMT). To obtain a broad view, we constructed deletions in the terS_ϕ genes of helper prophages 80α (a well-studied helper phage) and ϕNM1 (a recently identified helper phage) and tested them with a human (SaPI1 or SaPI2) or a bovine-derived (SaPIbov1 or SaPIbov2) SaPI, with and without deletions in their terS_SP genes (ΔterS_SP), all detoxified and marked with a tetracycline resistance gene (tetM). To score IMT frequencies, we introduced a standard laboratory non-conjugative rolling-circle-replicating vector (pOS1) into each of the test strains, in a phage- and SaPI-free derivative of S. aureus NCTC 8325 (RN450). The test strains were induced and the resulting lysates were tested for SaPI and plasmid transfer to S. aureus or L. monocytogenes recipients by selection for the corresponding antibiotic markers; we included L monocytogenes because we had previously observed intergeneric PMT from S. aureus to this species (Chen and Novick, 2009). As expected for classical PMT 80α ΦNM1 required TerS_Φ to transduce pOS1 (Figures 1A and 1E). Remarkably, when the genetically distinct and unlinked SaPI elements were introduced into strains with ΔterS_ϕ prophages, the ability to transfer pOS1 was restored (Figures 1B, 1F, and S2). Transfer of the SaPI element and IMT required TerS_SP, as ΔterS_ϕ/ΔterS_SP double mutants lost the ability to transfer either SaPI (Figure S1) or pOS1. Complementation with terS_SP restored the transfer defect in respective double mutants, in a promoter-inducible manner, confirming that SaPI element transfer and IMT required the TerS_SP terminases. Helper ΔterS_ϕ phages could not be complemented in trans for viral genome packaging because their pac sites are embedded in their structural terS_ϕ genes (Christie and Dokland, 2012; Ubeda et al., 2009). In contrast, the ΔterS_SP mutants were efficiently complemented for SaPI transfer in trans, suggesting that the SaPI pac sites are not embedded in the terS_SP genes. In addition to pOS1, SaPIs also transferred DNA elements native to S. aureus; namely, a naturally occurring non-conjugative theta-replicating plasmid that carries multiple antibiotic resistance genes (pI524) and the chromosomal accessory gene regulator (agr) locus, in a TerS_SP-dependent manner (Figures 1C, 1D, 1G, and 1H). These results reveal that SaPIs employ terminases to exploit phage packaging, forming transducing particles capable of intra- or intergeneric gene transfer of host DNA. IMT is distinct from classical PMT, as SaPIs are not defective phages, but are pathogenicity islands that are parasites of phages.

Figure 1. SaPIs Mediate Intra- and Intergeneric Gene Transfer of Plasmid and Chromosomal DNA.

Strains were MC-induced and the lysates tested for gene transfer to S. aureus (Sa) or L. monocytogenes (Lm). Genotypes are indicated as wild type (+), deletion (Δ), and absent (−). Strains were complemented with vector (V), terS_ϕ, or terS_SP under P_tet that was uninduced (U) or induced (I) with anhydrotetracycline. (A and E) 80α or ϕNM1 strains tested for PMT of vector pOS1, respectively. (B–C and F–H) 80α/SaPlbov1 or ϕNM1/SaPI2 strains tested for IMT of vector pOS1 (B and F), a naturally occurring plasmid pI524 (C and G), and a chromosomal DNA agr∷cadCA (D and H), respectively. The results are represented as transduction units (TRU) ml⁻¹. Values are means ± SD (n = 3 independent samples). See also Figures S1 and S2.

PMT and IMT do not Interfere With One Another

To reveal IMT, the phage terS_Φ genes were deleted to provide a direct demonstration of TerS_SP-dependent packaging of host DNA. However, we considered the possibility that IMT could be artifactual, owing to an increased availability of essential phage products, such as TerL and capsid (Figure 2A). To test for this possibility, we set up a test for IMT in the presence of the phage TerS_Φ. As small terminases determine DNA packaging specificity, we were able to distinguish TerS_Φ and TerS_SP activity on the basis of differential pac site recognition. Since homologs of phage pac sites exist throughout the bacterial genome, and are used at relatively high frequencies to package non-phage DNA (Chelala and Margolin, 1976; Schmieger, 1982), we reasoned that the same would be true for SaPI, and we designed a screen to isolate a clone from a chromosomal DNA fragment library (of a phage- and SaPI-free S. aureus strain) that would be differentially recognized by SaPI and enable direct measurements of IMT. From this screen, diagrammed in Figures S3 and described in the Supplemental Experimental Procedures (SEP) text, a library clone (pHFT-SP) that conferred much higher frequencies of IMT, than that of the vector alone, was selected for further analysis.

Figure 2.

SaPI-Mediated IMT Complements Phage-Mediated PMT.

(A) SaPI terminase and the phage DNA packaging machinery (TerS_ϕ, TerL, Portal).

(B–C) Strains were MC-induced and the lysates tested for plasmid transfer. Genotypes are indicated as wild type (+), deletion (Δ), and absent (−). The results are represented as transduction units (TRU) ml⁻¹. Values are means ± SD (n = 3 independent samples).

(B) PMT does not interfere with IMT. 80α or 80α/SaPlbov1 strains tested for IMT of vector (pJC1361) or pHFT-SP (pJC1552) to S. aureus.

(C) IMT does not interfere with PMT. 80α or 80α/SaPlbovl strains tested for PMT of vector (pJC1361) or pHFT-ϕ (pJC1551)to S. aureus.

(D) Electron micrograph of negatively stained 80α particles. Arrows indicate unused procapsids.

See also Figure S3 and Table S1.

Next, pHFT-SP was used to monitor TerS_SP-mediated packaging. When pHFT-SP was reintroduced to a RN450 (ΔterS_80α/terS_SPb1⁺) derivative, transfer of pHFT-SP was four orders of magnitude greater than that of the vector (Figure 2B), confirming that the DNA fragment carried on pHFT-SP confers high frequency IMT. In contrast, a terS_80α⁺/ΔterS_SPb1 single mutant transduced pHFT-SP at the same level as did the vector, indicating that TerS_80α is indifferent to the fragment carried on pHFT-SP. This result suggests that TerS_80α and TerS_SPb1 specificities do not overlap, and that the four orders of magnitude increase in transfer of pHFT-SP is specific and directly attributable to TerS_SPb1. In the wild type strain, when both small terminases were present, pHFT-SP was transferred at the same high frequency as when only the TerS_SPb1 was present. Parallel results were also obtained with ΦNM1/SaPIbov1 derivatives (Figure S3D). These results show that TerS_Φ does not inhibit TerS_SP DNA processing; IMT is significant and distinguishable from PMT, demonstrating that IMT is a natural phenomenon.

Conversely, we asked whether SaPI terminases inhibit PMT, by using the same chromosomal library to select for a high frequency PMT clone (pHFT-ϕ). As expected, pHFT-Φ was high frequency and specific for PMT, and retained this activity when both terminases were present (Figure 2C), indicating that TerS_SPb1 did not detectably diminish the transfer frequency of pHFT-ϕ by PMT. Parallel results were also observed for ϕNM1/SaPIbov1 derivatives (Figure S3D). Furthermore, in support of the generality of this mechanism for other SaPI-containing lysogens, the presence or absence of TerS_ϕ did not affect SaPI element transfer titers, nor did the presence or absence of TerS_SP affect phage titers (Table S1). These results suggest that IMT does not replace PMT, but complements it by forming transducing particles generated by a small terminase with disparate specificity, effectively expanding the repertoire of chromosomal regions that can be efficiently transferred.

The lack of phage and SaPI cross inhibition for DNA transfer is consistent with the principle that certain components of the phage DNA packaging machinery are not rate limiting for particle production. Phage small and large terminase genes often overlap or are adjacent to each other, such that they are co-transcribed and co-translated, resulting in TerS_Φ and TerL proteins at roughly equimolar ratios. Since the phage DNA packaging machinery utilizes TerS_Φ and TerL at approximately a 2:1 ratio, respectively, TerL is typically not limiting (Black, 1989; Casjens, 2011). Therefore, to determine if procapsids are limiting, we examined a phage 80α lysate by electron microscopy and observed that 9.7 ± 0.7% of the total phage heads exhibited the round and less geometric form that is characteristic of unused procapsids (Figures 2D and S3E) Given that 80α phage lysates are typically in the range of 1.0E+10 to 1.0E+11 plaque forming units ml⁻¹, the pool of free procapsids is large enough to account for all SaPI element transfer and IMT.

Uncoupling of IMT from the SaPI Life Cycle

To extend the above observations, we surveyed a series of prophage-SaPI combinations (Figure S1). In one combination, that of ϕNM1/SaPI2, we observed that the SaPI element transfer rate was several orders of magnitude lower than that seen with a typical helper phage (Figure S1K), showing that ϕNM1 is a sub-optimal helper phage for SaPI2. Remarkably, the transduction frequency for pOS1 was as high as that seen for strains with high frequency SaPI transfer (Figure 1F). A similar disparity was observed for ϕNM2, which is also a sub-optimal helper phage for SaPI2 (Figures S3F and S3G). These results indicated that IMT need not be obligatorily coupled to SaPI transfer.

In previous studies, we observed that SaPI operon 1, which includes terS_SP, is regulated by LexA and is therefore SOS inducible (Ubeda et al., 2007). DNA damage leads to activation of the SOS response, resulting in prophage repressor inactivation and prophage induction (Little, 1993). However, since the overall SaPI life cycle is not SOS-induced (Figure 3A), but rather is initiated by helper phage-mediated de-repression (Tormo-Mas et al., 2010), no obvious role for SOS induction of operon 1 could be envisioned at that time. Now, however, it seemed possible that induction of operon 1 could account for the above disparity: terS_SP would be expressed and mediate DNA packaging into ϕNM1 capsids, despite poor induction of the SaPI2 element. To test this, we used a non-helper prophage derived by deleting the de-repressor gene (dut) from phage 80α(Tormo-Mas et al., 2010), and tested it by comparing its ability to induce SaPIbov1 element transfer with its ability to induce IMT of pHFT-SP. As expected the Δdut 80α mutant generated wild type levels of phage, but was defective for SaPIbov1 element transfer (Figure 3B). In contrast, pHFT-SP was transduced at the same high frequency by the mutant as by the wild type 80α (Figure 3C), suggesting that the generation of plasmid transducing particles is independent of SaPI de-repression Because the Δdut 80α mutant is a derivative of a phage that is normally a SaPIbov1 helper, we repeated the experiment using SaPI1 and the natural non-helper phage ϕ11 (Lindsay et al., 1998; Tormo-Mas et al., 2010). Though the induction of ϕ11 prophage was unable to de-repress SaPI1 for element transfer (Figure 3D), the transfer frequency for pOS1 was comparable to the levels observed for SaPI1 with helper phage 80α (Figures 3E, S2B, and S2F), further demonstrating that SaPI element replication and transfer is not required for IMT.

Figure 3.

SaPI-Mediated Host Gene Transfer is Uncoupled from the SaPI Life Cycle and Does Not Require Specific Helper Phages.

(A) SaPI de-repression following SOS induction of SaPI helper or non-helper prophages.

(B–G) Lysates were tested for SaPI element-specific transfer (SPST), PMT, and IMT to S. aureus. Plaque forming units (PFU) were determined on S. aureus. Genotypes are indicated as wild type (+), deletion (Δ), and absent (−). The results are represented as PFU or SPST ml⁻¹. Values are means ± SD (n = 3 independent samples).

(B and C) Helper phage-coded SaPI inducer is not required for IMT following prophage induction. 80α/SaPlbov1 strains were tested for SPST and IMT of pHFT-SP (pJC1552), respectively.

(D and E) A natural non-helper prophage can be used for IMT. ϕ11/SaPI1 strains were tested for SPST and IMT of pOS1, respectively.

(F and G) Helper phage-coded SaPI inducer is not required for IMT following phage infection. SaPlbov1 non-lysogens were infected with 80α and the lysates tested for SPST and IMT of pHFT-SP (pJC1779), respectively. See also Figure S3.

Previous studies have shown that neither the induction of non-helper prophages nor the infection by non-helper phages, de-repress SaPIs for element transfer (Lindsay et al., 1998; Ruzin et al., 2001). To test for IMT with infecting non-helper phages, we generated non-lysogenic RN450 strains carrying SaPIbov1, infected them with the non-helper Δdut 80α mutant, and tested them by comparing SaPIbov1 transfer with the IMT of pHFT-SP. As shown in Figure 3F, the Δdut 80α mutant generated wild type levels of phage, but was defective for SaPIbov1 transfer. In contrast, pHFT-SP was transduced at the same high frequency by the mutant as by the wild type 80α (Figure 3G), suggesting that infecting phage induces SaPIbov1 terS expression. Thus, IMT is independent of SaPI de-repression during both induction and infection with non-helper phages.

Therefore, IMT can be uncoupled from SaPI self-transfer, lifting the narrow requirement for specific, rare helper phages, such that potentially any resident prophage or infecting pac phage can be used. Furthermore, these results reveal a fundamental distinction between phage- and SaPI-mediated host gene transfer: phage-mediated host gene transfer is an obligatory secondary by-product of viral maturation, while SaPI-mediated host gene transfer can be a primary SaPI function, independent of the SaPI life cycle.

Identification of SaPI Pseudo-pac Sites

In order to identify chromosomal regions that could be packaged and transferred by SaPIs at high frequencies, we scaled up the screen described above by testing putative high frequency IMT clones (~2,000) from the clone library in a micro-transduction format (Figure 4A). Putative clones were retransformed into a fresh RN450 (ΔterS_80α/terS_SPb1⁺) background, their transduction frequencies remeasured (Figures 4B and S4A), and the inserts of clones with transduction frequencies at least three orders of magnitude greater than of the vector were sequenced.

Figure 4.

Identification of SaPI Pseudo-pac Sites in the S. aureus Chromosome.

(A) Screen for ppac site clones. Lysates from S. aureus chromosomal DNA library clones (diagrammed in Figure S3), in a ΔterS_80α/terS_SPb1⁺ background, were diluted to 10⁻² and tested for IMT to S. aureus in a 48-well microtransduction format. Each well represents an individual clone. A scale from no transductants to high frequency transduction (HFT) is shown.

(B) HFT clones from the screen were retransformed into a fresh ΔterS_80α/terS_SPb1⁺ background and the resulting lysates tested over a dilution series to confirm high frequency IMT.

(C) SaPI ppac motifs.

(D) SaPI1/2 and SaPlbov1 pac sites. Arrows indicate direct repeats and a bracket indicates the variable region.

(E) The SaPI pac site. Lysates of ΔterS_80α/terS_SPb1⁺ strains carrying plasmids with a single GCTAAA repeat, ppac clone 22, a SaPI1 or SaPI2 pac site-containing fragment (SaPI1/2), a SaPI1/2 pac site corresponding to motif A (SaPI1/2-A), the SaPlbovl pac site-containing fragment (Bov1), the SaPI bov1 pac site corresponding to motif A (Bov1-A), and the shared motif B (B) were tested for IMT to S. aureus. The results are represented as transduction units (TRU) ml⁻¹. Values are means ± SD (n = 3 independent samples). See also Figure S5.

From this screen, we identified additional SaPI pseudo-pac (ppac) site-containing chromosomal fragments, and motif-based sequence analysis (MEME) of their sequences generated two complex overlapping motifs (A and B) with a discernible direct repeat (GCTAA[A/T], Figure 4C). To assess these motifs, we used them to identify the bona fide SaPI pac site, which was not yet known. Accordingly, we examined the SaPI genomes and found a matching direct repeat just upstream of operon I. In addition, a short variable 5′ region distinguishing SaPI1 and SaPI2 from SaPIbov1 (Figure 4D), corresponding to the sequence that distinguished motif A from B, was also noted. To determine whether this sequence element could serve as a SaPI pac site, we cloned the SaPI1/2 and SaPIbov1 regions encompassing the motifs, including flanking sequences, and tested them for IMT. Shown in Figures 4E and S5, we found that for all three SaPIs, the plasmid carrying the SaPI1/2 fragment was transferred at frequencies higher than the SaPIbov1 fragment, but plasmids carrying either fragment were transferred at higher frequencies than the SaPI ppac clone, indicating that all three SaPI terminases have similar DNA sequence recognition and confirming that this SaPI region, and not the terS_SP gene, contains the pac site. In contrast, a single copy of the repeat did not increase transduction frequency. To refine the SaPI pac site, we cloned SaPI fragments corresponding only to motifs A and B, tested them for IMT, and found that transfer frequencies for plasmids carrying motif A fragments were indistinguishable from the transfer frequencies of the corresponding larger fragments, indicating that sequences flanking motif A are not part of the SaPI pac site. When we tested the plasmid carrying the shared motif B sequence, the transfer frequency was indistinguishable from that of the SaPIbov1 fragments, indicating that the sequence in the variable region of the SaPI1/2 pac site, but not that in the SaPIbov1 pac site, is required for full terminase packaging activity for all three SaPIs. Therefore, motifs A and B likely represent close approximations of the SaPI1/2 and SaPIbov1 pac sites, respectively.

From the screen, we identified 22 individual SaPI ppac site-containing clones, and because we isolated numerous clones for every SaPI ppac site identified, we reasoned that either the screen had been saturated or a limitation of the DNA library had been reached. To estimate the number of ppac sites to be expected from a S. aureus genome, we applied motif B to 10 sequenced S. aureus genomes, using a motif alignment and search tool (MAST), and found an average of 20 matches (20.4 ± 2.5) per genome (Tables S2). To determine the number of sites expected by chance, we used a sequence randomizer (Randomseq) to generate 10 permutations of the NCTC 8325 genome, each with the same number of bases and G/A/T/C distribution, and found an average of 3 matches (2.9 ± 1.7) per randomized genome with motif B (Table S2). Furthermore, a second motif search tool (FIMO) yielded parallel results (Table S2). Therefore, these estimates suggest that S. aureus genomes contain significantly more matches to ppac site motif B than is predicted by chance.

To determine whether there are additional SaPI ppac sites, we searched the NCTC 8325 genome with motifs A and B (using MAST) and found 16 total matches with relatively strong p-values (<1.0e-06) that had not been previously identified. When we cloned these 16 and tested them for IMT, we identified an additional 6 SaPI ppac sites (Figure S4A). Since not all of the clones were high frequency for IMT, future mutational analysis of the SaPI pac sites will be required to further refine the ppac motifs. Moreover, when we cloned 16 additional alignments with weaker p-values (1.0e-07 to 2.0e-06), none was high frequency for IMT (Figure S6). Therefore, it is unlikely that weaker alignments with motifs A and B will reliably identify more, if any, additional SaPI ppac sites.

Using the methods described above, we identified 28 SaPI ppac sites in the S. aureus chromosome. A comparison of these sites from NCTC 8325 with those of multiple sequenced S. aureus genomes shows that most of them are identical in sequence, strand orientation, and relative location in the chromosome (Table S3), indicating that the SaPI ppac sites are highly conserved across different S. aureus strains. Taken together, these analyses suggest that many S. aureus genomes may be enriched for SaPI ppac sites.

Directionality of TerS_SP-Mediated DNA Packaging

When we applied the motifs to each ppac clone, a sequence corresponding to a putative SaPI ppac site could be identified (Table S4). To confirm these, we recloned the identified ppac sites from the high frequency IMT clones, tested them for IMT, and observed that they retained the full-scale high frequency IMT phenotype of their larger original clones (Figure S4B).

To determine the polarity of TerS_SP-mediated packaging, we chose a region of the S. aureus chromosome lacking nearby ppac sites – presumably a cold zone for SaPI-mediated IMT. We then constructed a test system in this region of the chromosome to determine packaging polarity, by inserting a non-replicating integration plasmid (next to SAOUHSC 00350), with or without the SaPI1/2 pac site, flanked by a cadmium resistance (Cd) cassette at a distance of 5 kb either upstream (Cd1) or downstream (Cd2) of the inserted pac site in a RN450 (ΔterS_ϕNM1/terS_SP2⁺) derivative (Figure 5A). A distance of 5 kb was chosen to include the drug resistance gene by headful packaging into a SaPI capsid, and leave sufficient flanking DNA for homologous recombination. These strains were then tested as donors of cadmium resistance to S. aureus. Shown in Figure 5B, when the motif was on the top strand, the frequency of Cd2 transduction was dramatically enhanced, but that of Cd1 was unchanged, relative to the frequency of transduction with the integrated vector. When the SaPI1/2 pac site orientation was flipped, such that the motif was on the bottom strand, the opposite was observed. These results show that the SaPI pac site directs packaging in the direction of operon I, toward the terS_SP gene in the SaPI genome. Furthermore, since pac sites are invariably asymmetric, owing to the unidirectional nature of terminase-mediated DNA packaging, the asymmetric SaPI pac site that we have identified, characterized by a direct repeat, is consistent with unidirectional packaging.

Figure 5.

TerS_SP-Mediated DNA Packaging is Unidirectional. (A) A test system to determine TerS_SP-packaging directionality. A non-replicating integration plasmid, with or without insert DNA, at the SaPI4 attachment site adjacent to SAOUHSC 00350. The test site is flanked by a Cd marker either 5 kb upstream (Cd1) or downstream (Cd2), in a ΔterS_ϕNM1/terS_SP2⁺ background. (B) Strains with DNA inserted next to SAOUHSC 00350 were tested for IMT of Cd1 or Cd2 to S. aureus. The SaPI ppac motifs are in the top strand of the SaPI1/2 pac, ppac 1, and ppac 23 clones, and the bottom strand of the SaPI1/2 pac-reversed (SaPI1/2-r), ppac 2, and ppac 27 clones. The results are represented as a ratio of the transduction units (TRU) ml⁻¹/TRU of the vector (TRUv) ml⁻¹. Values are means ± SD (n = 3 independent samples). See also Table S4.

Recently, a second sequence has been proposed as the pac site for SaPI1 (Bento et al., 2014) Thi 19 nt sequence overlaps with the last 11 nt’s of our minimal pac site (Figure 4D) and includes the second repeat, but does not match to ppac motif B, and it was postulated that a 16 nt inverted repeat (ACAAGCTAAAGTTTGTGTT) is actually recognized as the pac site. Although it is difficult to reconcile an inverted repeat with the observed unidirectional packaging of SaPI terminase (Figure 5B), it is possible this sequence may also contribute to pac site recognition.

To confirm that the same packaging directionality is followed for the SaPI ppac sites, we repeated the above test using four different high frequency IMT clones, two with the motif on the top strand and two with the motif on the bottom strand (Figure 5B). Both clones with the motif on the top strand enhanced transfer of Cd₂, but not of Cd1. Likewise, when the motif was on the bottom strand, both clones enhanced transfer of Cd1, but not of Cd2. Taken together, these results confirm that TerS_SP–mediated packaging is initiated by the recognition of ppac sites and proceeds from the 5′ to 3′ direction on the strand containing the ppac motif.

SaPI Pseudo-pac Sites Enhance the IMT Frequencies of Genes Associated with Disease

Figure 6A is a circular representation of the NCTC 8325 chromosome, with numbered arrows representing the identified SaPI ppac sites, showing that they are distributed throughout the genome and on both strands of the chromosome. Knowing the packaging polarity determined by the ppac sites, we then examined the gene content immediately downstream of each site, corresponding to the length of host DNA that would be included in the segment to be packaged by a headful packaging mechanism. Accordingly, we examined the gene content of the region immediately 16 kb downstream of each site (Tables S4 and S5), corresponding to the approximate length of a complete SaPIbov1 genome monomer including terminal redundancy (explained in SEP text). Listed in Table S5 and summarized in Figure 6A, 135 (31.3%) of the ppac-linked ORFs were identified as disease-associated genes involved in iron and phosphate acquisition, polysaccharide synthesis, secreted virulence, virulence, ion and pH homeostasis, and resistance mechanisms. By comparison, when we examined the gene content of the region immediately 16 kb upstream of each ppac site (not including regions overlapping with downstream regions), only 19 (6.4%) of the upstream ORFs were identified as disease-associated, which is similar to the overall NCTC 8325 genome at approximately 9% (Gillaspy et al., 2006). Therefore, segments of the chromosome that are linked to IMT have a much higher percentage of disease-associated genes than segments that are not linked.

Figure 6.

SaPI Pseudo-pac Sites are Linked to Disease-Associated Genes.

(A) The SaPI ppac sites mapped onto a circular representation of the NCTC 8325 chromosome. The outer and inner circles represent the relative top and bottom strands, respectively. SaPI attachment sites are indicated with a straight line. Arrows indicate ppac polarity of packaging, and ppac 5 and ppac 17 are repeated in the same direction. Inside the map is a summary of the disease-associated genes within a 16 kb headful limit immediately upstream (black) or downstream (blue) of the ppac sites (Tables S4 and S5).

(B–C) Cd markers were inserted into the chromosome downstream of native ppac sites in a ΔterS_ϕNM1/terS_SP2⁺ background and the strains were tested for IMT of the chromosomal Cd markers to S. aureus. Genotypes are indicated as wild type (+), deletion (Δ). The results are represented as transduction units (TRU) ml⁻¹. Values are means ± SD (n = 3 independent samples).

(B) PMT of chromosomal DNA downstream of native ppac sites. Cd markers were moved to RN450 (terS_ϕNM1⁺/ΔterS_SP2 or ΔterS_ϕNM1/ΔterS_SP2) as donors for PMT.

(C) IMT of chromosomal DNA downstream of native ppac sites. Cd markers were moved to RN450 (ΔterS_ϕNM1/terS_SP2⁺or ΔterS_ϕNM1/ΔterS_SP2) as donors for IMT.

Of the gene groups, iron acquisition genes had the most frequent linkage to IMT, with 13 of 28 ppac sites linked to known and predicted iron-related genes. Iron acquisition genes are functionally divided into host heme consumption and siderophore-mediated iron uptake, and the same dichotomy was reflected in the ppac-linked genes. Nearly all of the known heme-iron genes (Hammer and Skaar, 2011), from four unlinked loci around the chromosome, were ppac-linked, including the ferric uptake regulator (fur), the iron-regulated surface determinant (isdA-F) system, both heme-degrading monooxygenases (isdG and isdI), the heme-regulated transporter (hrtAB), and the heme sensor system (hssRS). In contrast, many of the siderophore-mediated iron uptake genes were not found to be ppac-linked, indicating a possible functional distinction between iron-acquisition genes by IMT.

Among the other ppac-linked determinants were all of the known genes involved in inorganic phosphate (P_i) sensing and uptake, including the phosphate-specific transporter (pstSABC) and the two-component phosphate starvation system (phoRP) (Bergwitz and Juppner, 2011); most of the genes involved in exopolysaccharide synthesis, including the capsular polysaccharide (cap5A–C, E–P) genes involved in resisting phagocytosis (O’Riordan and Lee, 2004) and the polysaccharide intercellular adhesion (icaRADBC) genes involved in biofilm formation (O’Gara, 2007); many virulence genes, including two major global virulence regulators (sarA and arlR) (Cheung et al, 2008) and the phenol-soluble modulins (psma1–4) (Peschel and Otto, 2013); and a large number of genes involved in ion and pH homeostasis and resistance. Furthermore, we observed that some ppac sites were paired in close proximity and oriented so that their packaged segments overlap (Figure 6A), such that together they effectively covered a continuous local region of the chromosome that is much larger than can be accommodated in a 16 kb packaging limit. In some pairings, ppac sites such as 4 and 5 can combine to cover all of the superantigen-like (ssl) genes, while ppac sites 12 and 13 can combine to transfer all of the isd genes at one locus, suggesting that paired ppac sites may be coordinated for the transfer of large regions containing sets of related genes.

To confirm the packaging of disease-associated genes directed by naturally occurring ppac sites, we inserted a Cd marker into the chromosome 12 kb downstream of the native ppac 2 site in the cap5 operon, 12 kb downstream of the ppac 12 site in the isd operon, and 5 kb downstream of the ppac 16 site in the pst operon. In addition, to show that some ppac sites can coordinate to effectively transfer a larger region of the chromosome, a Cd marker was inserted 10 kb downstream of the ppac 22 site (but upstream of the ppac 23 site) and 10 kb downstream of the ppac 23 site (but upstream of the ppac 22 site), since ppac 22 and ppac 23 face each other and form an overlapping pair. These Cd markers were moved to ϕNM1/SaPI2 derivatives of RN450, tested for Cd marker transfer to S. aureus, and compared to the cadmium resistance transduction frequency with that of markers upstream (Cd1) or downstream (Cd2) in the absence of any vector inserted at SAOUHSC 00350. Shown in Figure 6B, PMT of Cd markers linked to SaPI ppac sites in the terS_ϕNM1⁺/ΔterS_SP2 background were not significantly greater than that of either Cd1 or Cd2 in most cases, except for the ppac 12-linked insertion in the isd operon, suggesting that ϕNM1 ppac sites are not linked to most of these regions in the chromosome. In contrast, IMT of Cd markers linked to ppac sites were significantly greater than that of either Cd1 or Cd2 in the ΔterS_ϕNM1/terS_SP2⁺ background (Figure 6C), indicating that chromosomal regions linked to SaPI ppac sites are packaged and transferred at higher frequencies by IMT. Furthermore, ppac sites that face each other and form an overlapping pair, such as ppac 22 and ppac 23, can effectively transfer a larger region of the chromosome than can be accommodated in a SaPI small capsid.

Next, to determine whether IMT-mediated gene exchange could result in a phenotypic change in the recipient, we set up a system to assay for the transfer of unmarked chromosomal DNA. Using rpsL-based counterselection (explained in SEP text), a cassette containing a wild type copy of the rpsL gene was inserted into each of the SaPI ppac-linked genes (that were previously marked with a Cd cassette above) of a streptomycin resistant derivative of the S. aureus strain Newman, generating a series of merodiploid knockin (KI) strains that are no longer viable on streptomycin; therefore, only KI strains that have lost the rpsL⁺ cassette, either by allelic exchange or spontaneous excision, can grow on streptomycin. Here, we switched to a Newman background because not all of the genes with KIs are functional in NCTC 8325. Representative schematics of the ppac 12 and ppac 2 KIs are shown in Figures 7A and 7B, respectively. Next, a ΔterS_ϕNM1/terS_SP2⁺ derivative of TB4 (a phage- and SaPI-free variant of Newman) was generated as an IMT donor strain and tested for the transfer of unmarked wild type DNA to the KIs with selection on streptomycin. As expected, and shown in Figure 7C, reversion of the KIs to streptomycin resistance paralleled the frequencies previously observed for the transfer of the corresponding Cd markers, and were again significantly greater than that of either insertion upstream (rpsL1) or downstream (rpsL2) of SAOUHSC 00350. To confirm that the KI revertants were not second site suppressor mutants, PCR analysis on the chromosomal DNA of the isdC or cap5H KI revertants confirmed that the transductants had lost the insertion (Figures 7E and 7F).

Figure 7.

IMT Mediates the Allelic Exchange of Disease-Associated Genes Linked to SaPI Pseudo-pac Sites.

(A) Diagram of isdC KI in Newman (NM).

(B) Diagram of cap5H KI Newman (CP5 type) or cap8H KI in MN8 (CP8 type). (C–D) A TB4 (ΔterS_ϕNM1/terS_SP2⁺) derivative was tested as a donor for IMT of unmarked chromosomal DNA to KI strains. The results are represented as transduction units (TRU) ml⁻¹. Values are means ± SD (n = 3 independent samples).

(C) IMT of Newman DNA to KIs in a Newman rpsL* background.

(D) IMT of Newman DNA to KIs in a MN8 rpsL* background.

(E–G) Confirmation of IMT-mediated Kl reversion. Size marker is a 1 kb ladder (NEB). (E) PCR with isd-specific primers (JCO 692 + JCO 694). (F) PCR with cp5-specific primers (JCO 700 + JCO 702). (G) PCR with cp5-specific or cp8-specific primers (JCO 700 + JCO 706).

(H) IMT-mediated KI reversion results in a phenotypic change. Iron starved strains were grown in iron-chelated RPMI ± lysed red blood cells (LRBC). Strains are indicated as wild type (WT), isd KI (KI), isd KI transductant (TR). The results are represented as measurements of OD₆₀₀. Values are means ± SD (n = 3 independent samples). See also Figure S7.

Because KI reversion could have resulted from either allelic exchange or spontaneous excision, we determined whether IMT could mediate a capsular polysaccharide (CP) type switch to distinguish between these two possibilities. Diagrammed in Figure 7B, most of the gene clusters for CP are nearly identical, except for the genes that determine CP type, which are located in the central region and show very little homology between the two loci (O’Riordan and Lee, 2004). Therefore, reversion of a KI in any of the specificity genes by allelic exchange will result in a CP type switch, while spontaneous excision will retain the original CP type. Accordingly, a cap8H KI was generated in a streptomycin resistant derivative of the CP8 type S. aureus strain MN8 (Figure 7B), and this strain was used as a recipient for IMT with lysate from the TB4 (ΔterS_ϕNM1/terS_SP2⁺) derivative (CP5 type) from above. As expected, the reversion frequency of the cap8 KI was significantly greater than that of either rpsL1 or rpsL2 (Figure 7D). PCR analysis of the cap8 KI revertants confirmed that not only had the transductants lost the insertion, but they had also switched from CP8 to CP5 (Figure 7G), showing that reversion was the result of allelic exchange.

To confirm a phenotypic change in the revertants, we tested the isdC KI transductants for the functional restoration of the Isd system (Hammer and Skaar, 2011), which is responsible for capturing nutrient iron from heme (explained in SEP text). The isdC KI has a polar insertion in the first gene of the isdCDEFsrtBisdG operon (Figure 7A), and is predicted to exhibit a severe loss-of-function phenotype for heme acquisition (Mazmanian et al., 2003). On the basis that iron is required for S. aureus growth, iron-starved cultures were used to inoculate iron-chelated media and bacterial growth was measured by optical density. Shown in Figure 7H, all three strains (wild type, isdC KI, and isdC KI transductants) were unable to grow in iron-chelated media, but did not exhibit a growth defect in unchelated media (Figure S7). When the same iron-chelated media was supplemented with heme, the isdC KI was still unable to grow, but growth of the isdC KI transductant was comparable to wild type (Figure 7H), indicating that the isdC KI transductant had been functionally restored to wild type.

The above results suggest that ppac-linked genes can be exchanged between strains more frequently (than unlinked genes) by IMT. Furthermore, since the headful packaging mechanism does not distinguish between DNA segments downstream of a ppac site, the transfer of wild type genes or mutant alleles, of say adaptive mutations, would be equally likely. Taken together, these results suggest that the SaPI ppac sites may not be randomly scattered throughout the genome, as they appear to be located near chromosomal regions containing genes associated with adaptation to environmental contingencies, including many genes associated with S. aureus pathogenicity.

Discussion

Facultative pathogenic bacteria exhibit remarkable genetic plasticity, often depending on the horizontal transfer of pathogenicity islands, facilitating the rapid evolution of virulent and antibiotic resistant strains. The SaPIs are exemplary in this respect, having an extremely efficient mobilization mechanism, which enables them to transfer their self-coded virulence genes at high frequencies. In this report, we describe the discovery and in-depth documentation of a mode of horizontal gene transfer by the SaPIs, above and beyond the transfer of the islands themselves. This mode of transfer depends on recognition by the SaPI terminase of SaPI ppac sites occurring throughout the host cell’s genome. It occurs along with SaPI maturation, and is analogous to phage-mediated generalized transduction, which is coupled to phage maturation. It also occurs in the absence of SaPI maturation, and involves the independent activation of terS_SP expression, owing to the independent inducibility of SaPI operon 1 which encodes the SaPI terminase. This operon is expressed during prophage induction or phage infection, and results in SaPI terminase-mediated packaging of the same host genomic segments as those packaged during helper phage-mediated SaPI induction, and at essentially the same frequencies.

Uncoupled from the SaPI life cycle, IMT remains a function for the host for as long as the SaPI element is retained; and in that sense, it is analogous to the more classical functions (toxins and virulence factors) coded by pathogenicity islands. A dedicated mode of gene transfer that is induced during all phage interactions can be extremely advantageous, as phages are the most abundant biological entities in the biosphere, and are estimated to destroy up to 40% of all bacteria each day (Hendrix et al., 1999). Based on these estimates, most of the interactions between S. aureus strains and phages are predicted to be lytic infections by environmental phages, often fleeting, and seldom repeated. Even prophages, the residents of bacterial chromosomes, are but “transient passengers” in most strains, as their genomes are frequently subjected to an ongoing decay process that results in their inactivation and eventual elimination or replacement (Brussow et al., 2004). Since phages are extraordinarily diverse, presumably with a corresponding variety of small terminase specificities, they provide a highly irregular sampling of the host genome for genetic exchange. We have analyzed 3 different SaPIs that happen to have a common minimal pac site, and have found that the corresponding ppac sites are often positioned to promote the packaging of chromosomal regions that contain genes important for adaptation to metazoan hosts. Therefore, to counter the unpredictable gene transfer by the extant staphylococcal transducing phages, we propose that IMT directs transducing phages to transmit a set of specific gene segments that are linked to SaPI ppac sites. The SaPIs, of course, will most often remain quiescent while promoting the transfer of these particular host genomic segments.

Phages typically have narrow host ranges, and phage-mediated gene transfer is considered restricted to closely related bacteria. Likewise, SaPI element transfer requires specific phages, and is, by default, limited to the same narrow host ranges of their helper phages. By comparison, IMT can presumably be activated by any phage, and host gene transfer is independent of the rare SaPI helper phages. Given that SaPIs are only known to modify phage capsid size, and provided that the phage large terminase and SaPI terminase are compatible, IMT can potentially assume the tail specificity of any transducing phage for a seemingly unlimited host range.

At the level of single bacterial cells, it is difficult to imagine how SaPI ppac sites could have evolved. A cell in which a DNA sequence could have mutated into a novel ppac site would have, by necessity, been lysed by the time the recipient had received the DNA. However, bacterial strains typically do not function as individuals, but as communities of rapidly dividing cells with a collective purpose, and a novel ppac site would likely distribute to many siblings before a lytic event could take place. Alternatively, the SaPI terminases could have evolved to recognize a DNA signature that was cryptically associated to certain genes of like function. Thus, any SaPI-carrying strain with ppac sites linked to genes important for adaptation to a niche, say of a host, would presumably gain a selective advantage. Naturally, it is predictable that similar linkages to phage ppac sites also exist, but as mentioned above, most prophages are only temporary residents of bacterial chromosomes, and these associations may be short-lived on an evolutionary timescale. Therefore, perhaps IMT-linked ppac sites have had lasting selective value during evolution, linking the SaPIs to a much broader repertoire of virulence determinants than can be carried by the SaPIs themselves, creating a genetic “cloud” for the storage of virulence determinants.

Experimental Procedures

Additional experimental procedures are available online in the Supplemental Information.

Bacterial Strains and Growth Conditions

Bacterial strains were grown as previously described (Chen and Novick, 2009; Novick, 1991). The bacterial strains used in this study are listed in Table S6.

DNA Manipulations

The plasmids and oligonucleotides used in this study are listed in Table S6.

Transductions

For lysates, strains were grown to mid log, diluted to OD₆₀₀ = 0.5, and adjusted to 2 μg ml⁻¹ Mitomycin C (Sigma) or infected with an MOI of 0.1 until complete lysis. Lysates were then adjusted to 1 μg ml⁻¹ DNAse I + 1 μg ml⁻¹ RNAase and filter sterilized (0.2 μm pore) before use. For transductions, cells were infected for 30 minutes and then adjusted to 100 mM sodium citrate, mixed with 3 ml of top agar, and plated on selective agar.