Report

Phage-Mediated Intergeneric Transfer of Toxin Genes

See allHide authors and affiliations

Science  02 Jan 2009:
Vol. 323, Issue 5910, pp. 139-141
DOI: 10.1126/science.1164783

Abstract

Because bacteriophages generally parasitize only closely related bacteria, it is assumed that phage-mediated genetic exchange occurs primarily within species. Here we report that staphylococcal pathogenenicity islands, containing superantigen genes, and other mobile elements transferred to Listeria monocytogenes at the same high frequencies as they transfer within Staphylococcus aureus. Several staphylococcal phages transduced L. monocytogenes but could not form plaques. In an experiment modeling phage therapy for bovine mastitis, we observed pathogenicity island transfer between S. aureus and L. monocytogenes in raw milk. Thus, phages may participate in a far more expansive network of genetic information exchange among bacteria of different species than originally thought, with important implications for the evolution of human pathogens.

A major class of bacteriophages, the transducing phages, can package bacterial host DNA and transfer it to new hosts. The frequency of this transfer is dramatically increased for genetic elements that are designed for packaging by phages. At the same time, the ability of a bacterial species to intercept and sample the phage-carried genetic information of foreign species and genera could represent a vast source of new genetic information. However, phages are generally assumed to have narrow host ranges (1, 2), and phage-mediated genetic exchange is considered to be restricted. Given the numbers of extant bacteria and phages, intergeneric phage-mediated genetic exchange is a possibility. Of particular concern is the potential for the phage-mediated transfer of virulence determinants (3).

Highly mobile toxin-carrying pathogenicity islands of Staphylococcus aureus (SaPIs) are specially adapted to packaging and transfer by particular staphylococcal phages. The SaPIs, which encode toxic shock toxin (TSST-1) and other superantigens, are 14- to 18-kb elements that are inserted at specific chromosomal sites and are induced to excise and replicate by certain temperate phages. After replication, SaPI DNA is packaged into special small infectious particles in numbers similar to those of the plaque-forming phage particles, resulting in extremely high SaPI-specific transfer frequencies (4, 5). Although they are ubiquitous among S. aureus strains (6), the SaPIs are extremely rare in non-aureus staphylococci and have yet to be described in other bacterial genera (7), despite their great mobility.

We analyzed the sequence specificity of the SaPI1 insertion site by making an attC deletion, which was then used as the recipient for the transduction of a detoxified derivative of SaPI1 marked with tetracycline resistance (tetM). Although we expected SaPI1 transfer to the ΔattC mutant to be rare, we found that phage 80α (4) transferred SaPI1 to the ΔattC strain at nearly the same frequency as to the parental strain, with an intact attC site (Table 1). Southern blots of the ΔattC transductants indicated single integrations in a variety of secondary sites (fig. S1). Sequencing of SaPI1 insertion junctions (Fig. 1) confirmed that all the secondary integrations involved single crossovers with the SaPI1 element attS site, indicating that the secondary insertions were integrase-mediated events. An SaPI1 mutant lacking integrase could not be transduced either to the attC+ strain (8) or to the ΔattC strain (table S1).

Fig. 1.

Alternative SaPI1 insertions in S. aureus and L. monocytogenes. The chromosomal locations of SaPI1 insertions are taken from the sequenced genomes of S. aureus 8325 and L. monocytogenes EGDe. The wild-type primary SaPI1 attC is precisely duplicated upon integration and is indicated in bold. The left and right insertion junctions (JL and JR) are underlined and they indicate the hybrid SaPI1 attachment sites created from integration into a secondary attachment site. Red boxes indicate mismatches with the wild-type primary SaPI1 attC. Gray boxes indicate mismatches with the flanking chromosomal regions. S. aureus SAOUHSC_00844* is the chromosomal location of the wild-type primary SaPI1 attC. ORF, open reading frame.

Table 1.

High-frequency transfer of S. aureus pathogenicity islands SaPI1 and SaPIbov1 to L. monocytogenes. TRU, TRU/ml; PFU, PFU/ml on RN450. Sa, S. aureus. Lm, L. monocytogenes. JCSA109, SaPI1 ΔattC. Strains were transduced with S. aureus helper phages 80α, Φ11, ΦNM2, or ΦNM4 carrying SaPI1, SaPIbov1, or pRN0870 (pT181 replicon). There were no TRU observed for cell- or phage-only controls. The results are represented as the ratios of TRU to PFU/ml. %PA, relative phage adsorption, where 0% = no phage bound. Values are means ± SD (n = 3 independent samples). NO, none obtained at the highest phage titer.

View this table:

Alignment of the secondary sites revealed that they bore only a modest resemblance to the primary attC site; however, a cytosine pair at the putative recombination crossover point was always present (fig. S2). These results implied that SaPI1 integrase had much lower sequence specificity than other typical integrases, such as that of coliphage lambda, where transduction frequencies are reduced by more than 200-fold in the absence of the primary attachment site (9). The frequency and variability of secondary sites in S. aureus increase the probability that suitable insertion sites are present in many bacterial species, and any bacterium that can adsorb SaPI helper phage is a potential recipient.

We tested a diverse set of Gram-positive bacterial species as recipients for transduction with a detoxified derivative of SaPI1 (SaPI1-tst::tetM). Rare SaPI transduction to S. epidermidis and S. xylosus had been reported previously (7) and was confirmed for S. xylosus (table S2). With the striking exception of Listeria monocytogenes, transfer to other genera was not detected (table S3). Not only was SaPI1 transferred to L. monocytogenes efficiently, but also at the high frequencies observed with S. aureus (Table 1). Of the representative strains of L. monocytogenes serotypes that we tested, only those of serotype 4b were not transduced. Because the sequenced genomes of L. monocytogenes do not appear to have attC sites of SaPI1 or SaPIbov1 or even close variants, integration in secondary attachment sites was most likely. Sequencing of SaPI1 insertion joints in transductants of strain EGDe identified chromosomal L. monocytogenes DNA flanking integrated SaPI1 at several sites (Fig. 1 and fig. S3), confirming that the alternative SaPI attachment sites in L. monocytogenes are similar to those in S. aureus and that the SaPI1 integrase catalyzes these insertions [supporting online material (SOM) text].

To evaluate the generality of SaPI transfer to L. monocytogenes, we tested other mobile genetic elements and found that several were transferred efficiently. An additional SaPI, the genotypically distant SaPIbov1 (10, 11), and a detoxified derivative (SaPIbov1-tst::tetM) (10) gave results similar to those with SaPI1 (Table 1). Furthermore, SaPI mobilizing (helper) phages φ11 (12), φNM2, and φNM4 (13) transferred SaPI elements as efficiently as 80α (Table 1).

Standard phage susceptibility tests are based on plaque formation, but none of the four phages formed plaques on any of the seven serotypes of L. monocytogenes tested (table S4). Thus, the true overall host range of a phage may be much wider if it includes infection without plaque formation, which can be assessed only by gene transfer or phage DNA delivery. To distinguish whether the variation in transduction frequency was due to phage adsorption or host DNA restriction, we measured helper phage adsorption by L. monocytogenes. Serotypes 1/2a and 3b exhibited phage adsorption frequencies comparable to that of the staphylococcal control and were the strains most amenable to transduction, indicating there was a correlation between phage adsorption and transfer. This also implied that DNA restriction was not an important determinant of transduction frequency for these serotypes.

We then tested for the transfer of a virulence determinant unrelated to the SaPIs and used a detoxified derivative of phage φSLT (14) containing a tetracycline-resistance marker (tetM) inserted into the Panton-Valentine leukocidin (PVL) locus (15). Although ΦSLT generates lysates with relatively low titers, its transfer to L. monocytogenes was demonstrated by selection for the tetM marker (table S5). In contrast to SaPI transfer, this transduction requires lysogenization because PVL is carried in the φSLT genome. Similarly to the four SaPI helper phages, ΦSLT did not form plaques on L. monocytogenes (table S4), and converted strains did not liberate detectable plaque-forming phage particles upon mitomycin C induction. The potential for environmental ΦSLT transduction to L. monocytogenes is disconcerting, considering that PVL has been implicated in staphylococcal diseases (15). All the staphylococcal phages we tested mediated genetic transfer to L. monocytogenes.

We predicted the occurrence of phage-mediated SaPI transfer in an environment in which S. aureus and L. monocytogenes occur together. These species are common causes of bovine mastitis (1618), and we analyzed cow's milk as a medium for spontaneous prophage induction and SaPI1 transduction. When detoxified S. aureus derivatives of laboratory and clinical isolates were co-cultured with L. monocytogenes strains in raw milk, we detected spontaneous prophage induction and transfer of SaPI1 and SaPIbov1 to L. monocytogenes (SOM text and table S6).

Bovine mastitis costs the world's dairy industries billions in revenue each year; roughly 11% of total production is lost annually (19). Of the mastitis pathogens, S. aureus is of particular concern because of the low cure rate with antibiotic treatment and the rapid rise of antibiotic-resistant strains (20). A promising alternative to antibiotic treatment of S. aureus infections is phage therapy, which is currently the focus of several clinical trials for bovine mastitis (2124). To determine whether this strategy could promote SaPI transfer, we co-cultured S. aureus strains carrying detoxified SaPIs with streptomycin-resistant derivatives of L. monocytogenes strains EGDe and SK1442 in raw milk, adding SaPI-less phage 80α (propagated on RN450). As expected, high titers of phage 80α were efficient at clearing the S. aureus strains (table S7); however, the phage particles resulting from lysis also promoted the transfer of SaPI1 and SaPIbov1 to L. monocytogenes (Fig. 2).

Fig. 2.

Phage treatment of S. aureus promotes transfer of SaPI to L. monocytogenes in milk. S. aureus wild-type RN6734 carrying SaPI1 or SaPIbov1, clinical isolate RN4282 (RN6938) carrying SaPI1, or clinical isolate RF122 carrying SaPIbov1 (RN9749) were co-cultured with streptomycin-resistant variants of L. monocytogenes strains EGDe and SK1442 and phage 80α [107 plaque-forming units (PFU)/ml] in milk for 3 hours at 30°C. The samples were adjusted to 100 mM sodium citrate and plated for SaPI transductants, with selection against the donor strains. The cryptic phages of RN4282 and RF122 do not confer resistance to phage 80α. TRU, transduction units. “+” indicates the addition of phage. The results are represented as TRU/ml. There were no TRU observed for “–“RN6734 samples and phage-only controls. The “–“RN4282 and RF122 samples were below 100 TRU/ml. Values are means ± SD (n = 3 independent samples). Results and units for EGDe-S and SK1442-S are in gray and black, respectively.

Although superantigen-producing L. monocytogenes strains have not yet been reported, it is certainly true that environmental isolates of S. aureus carrying SaPIs are ubiquitous. Thus, the widespread use of anti-aureus phages in agriculture may accelerate the spread of staphylococcal toxins to Listeria or to any other bacteria to which the phages can adsorb (SOM text).

Phages form the framework for a living network of genetic information, interconnecting the microbes of the biosphere. This study hints that there is a pipeline of silent phage-mediated genetic information transfer among bacteria, indicating that phages are involved in far more numerous microbial connections than previously imagined.

Supporting Online Material

www.sciencemag.org/cgi/content/full/323/5910/139/DC1

Materials and Methods

SOM Text

Figs. S1 to S4

Tables S1 to S9

References

References and Notes

View Abstract

Navigate This Article