Inducible prophages contribute to Salmonella virulence in mice


Lionello Bossi. E-mail; Tel. (+33) 1 69 82 31 37; Fax (+33) 1 69 82 32 30.


We show that Salmonella typhimurium harbours two fully functional prophages, Gifsy-1 and Gifsy-2, that can be induced by standard treatments or, more effectively, by exposing bacteria to hydrogen peroxide. Curing bacteria for the Gifsy-2 prophage significantly reduces Salmonella's ability to establish a systemic infection in mice. Cured strains recover their virulence properties upon relysogenization. Phage Gifsy-2 carries the sodC gene for a periplasmic [Cu,Zn]-superoxide dismutase previously implicated in the bacterial defences against killing by macrophages. The contribution of the Gifsy-1 prophage to virulence — undetectable in the presence of Gifsy-2 as prophage — becomes significant in cells that lack Gifsy-2 but carry the sodC gene integrated in the chromosome. This confirms the involvement of Gifsy-2-encoded SodC protein in Salmonella pathogenicity and suggests that the Gifsy-1 prophage carries one or more additional virulence genes that have a functional equivalent on the Gifsy-2 genome.


Since the original description of toxin-encoding bacteriophages of Corynebacterium diphtheriae, evidence linking phages to bacterial pathogenicity has accumulated (reviewed by Barksdale and Arden, 1974; Cheetham and Katz, 1995; Waldor, 1998). In Salmonella enterica, the proximity of virulence genes to bacteriophage genes (Pulkkinen and Miller, 1991; De Groote et al., 1997; Farrant et al., 1997; Valdivia and Falkow, 1997; Gunn et al., 1998; Hardt et al., 1998) or to sequences reminiscent of phage attachment sites (Blanc-Potard and Groisman, 1997; Hensel et al., 1997; Wood et al., 1998) suggests that bacteriophages have contributed — and possibly still do — to the lateral spreading of pathogenicity determinants in this organism and related species. Yet, lack of direct evidence for a transfer mechanism as well as the apparent defective nature of the phage sequences involved (Hardt et al., 1998) have made phage involvement in the evolution of Salmonella pathogenicity remain speculative.

In a previous study, we described the existence of two prophage-like elements in the chromosome of Salmonella enterica serovar Typhimurium (hereafter referred to as S. typhimurium ). The two elements, named Gifsy-1 and Gifsy-2, are located at 57 and 24 units of the Salmonella genetic map, respectively, and share sequence homology over a portion of their DNA (Figueroa-Bossi et al., 1997). Their relatedness was also suggested by the finding that a Gifsy-1 regulatory mutation induces an activity causing frequent excision of both putative prophages. In the mutant background, DNA excision is not followed by replication; thus, the excised material is rapidly lost. Intriguingly, the frequency of Gifsy-1 loss is much higher in cells that contain Gifsy-2 compared with cells that have lost it, suggesting that Gifsy-2 somehow contributes to the excision reaction. Although the basis for this complementation remains unknown, the phenomenon allowed us to isolate derivatives of strain LT2 cured for Gifsy-1 or both Gifsy-1 and Gifsy-2 elements (Figueroa-Bossi et al., 1997).

During the course of this work, we realized that both putative prophages carried a sequence reportedly transcriptionally activated during Salmonella adaptation to oxidative stress (Wong and McClelland, 1994). These findings led us to examine the possibility that the Gifsy elements influenced some aspects of Salmonella interaction with its animal host. In parallel with this analysis, we examined the functional status of the two elements. The data presented below show that both Gifsy elements can be released as phages that are capable of relysogenizing formerly cured recipient strains. Such a release occurs spontaneously at low levels and can be stimulated by various treatments, the most effective of which involves exposing bacteria to hydrogen peroxide. We show that the Gifsy-2 prophage contributes significantly to Salmonella pathogenicity in mice and, once released as phage, it can transfer its virulence-conferring traits to formerly cured attenuated strains. The Gifsy-2 genome includes the sodC gene coding for a periplasmic superoxide [Cu,Zn]-dismutase previously implicated in Salmonella virulence (De Groote et al., 1997; Farrant et al., 1997). The contribution of the Gifsy-1 prophage to virulence is insignificant in the presence of Gifsy-2 as prophage, but can be revealed upon removal of the latter, provided that the sodC gene is introduced in the genetic background.


Gifsy-cured strains are attenuated in mice

The presence of the Gifsy prophages was examined in three S. typhimurium strains commonly used for virulence studies: ATCC14028s (Fields et al., 1986), SL1344 (Hoiseth and Stocker, 1981) and C52 (Coynault et al., 1996). Molecular and genetic tests demonstrated that all three strains contain both prophage elements at the same chromosomal locations as in strain LT2 (Fig. 1; data not shown). Strains were cured for the prophages using a mutation that destabilizes both elements (sbcE21 ) initially described in LT2 (Figueroa-Bossi et al., 1997; see Experimental procedures ). For all three strains, we were able to obtain isogenic derivatives lacking Gifsy-1 or both Gifsy-l and Gifsy-2 prophages (Fig. 1). These strains were used for mouse virulence assays. In an initial series of experiments, bacteria were administered orally. These experiments (Table 1) showed that the removal of both prophages leads to a significant attenuation of virulence in all three strains. Comparing the lethality of derivatives cured for Gifsy-1 alone (MA5973) or for both prophages (MA5975) suggested that the contribution to virulence results largely from the Gifsy-2 prophage (Table 1). This point was addressed further by using ATCC14028s-derived strains delivered intraperitoneally. The results (Table 2) confirmed the strong attenuation of the doubly cured strain and the lack of a significant contribution by the Gifsy-1 prophage to these effects. As a more sensitive test for the role of Gifsy-1, we compared the persistency of strains carrying or lacking this prophage in mouse spleens after mixed infections. Two experiments were performed. In the first experiment, five mice were challenged with an equal mixture of bacteria from a wild-type (ATCC14028s) and from a Gifsy-1-cured strain [MA5973; about 50 colony-forming units (cfu) from each strain injected intraperitoneally]. Five days later, mice were sacrificed and their spleens removed and analysed. In all cases, organs were found to contain approximately equal numbers of bacteria from both strains. In a second experiment, five mice were injected with the doubly cured strain (MA5975) and a strain lysogenic for Gifsy-1 (MA6275). About 1000 cfu from each strain were injected. This time, more than 90% of bacteria recovered from spleens 5 days later were lysogenic for Gifsy-1 (10 bacteria tested from each spleen). These data suggest that the Gifsy-1 prophage is completely dispensable for virulence when the Gifsy-2 prophage is present, but it plays a distinct role in the absence of the latter.

Figure 1.

. Pulsed-field gel electrophoretic separation and Southern analysis of XbaI-digested chromosomal DNA from virulent S. typhimurium strains and their Gifsy prophage-cured derivatives. The hybridization probe was a 32P-labelled plasmid pK4114 DNA (see Experimental procedures ). This probe hybridizes to a Gifsy-1-specific fragment (≈35 kb or 65 kb, depending on the strain), to a Gifsy-2-specific fragment (in the 800–850 kb range) and to the fusion fragment resulting from Gifsy-1 excision (≈200 kb; indicated by an arrowhead). Lanes 1–3: DNA from ATCC14028s-derived strains. The parental strain (lane 1) and derivatives cured for Gifsy-1 (lane 2) or for both Gifsy-1 and Gifsy-2 (lane 3). Lanes 4–6: DNA from C52-derived strains The parental strain (lane 4) and derivatives cured for Gifsy-1 (lane 5) or for both Gifsy-1 and Gifsy-2 (lane 6). Lanes 7 and 8: DNA from SL1344-derived strains. the parental strain (lane 7) and a derivative cured for both Gifsy-1 and Gifsy-2 (lane 8).

Table 1. . Mouse virulence properties of wild-type and prophage-cured strains of S. typhimurium. a. Mice were inoculated orally as described in Experimental procedures.Thumbnail image of
Table 2. . Mouse virulence properties of prophage-cured and relysogenized derivatives of S. typhimurium ATCC14028s. a. Mice were inoculated intraperitoneally as described in Experimental procedures.b. Strain MA6275 is a strain MA5975 derivative relysogenized by Gifsy-1.c. Strain MA6309 is a strain MA6275 derivative relysogenized by Gifsy-2.Thumbnail image of

The Gifsy prophages are active viruses

In the initial stages of this work, we observed that cultures of ATCC14028s released particles forming tiny plaques on lawns of doubly cured strain MA5975. From such plaques, we isolated cells singly relysogenized by Gifsy-1 (MA6275). Genetic and physical tests demonstrated that the new prophage was located at its original chromosomal position (data not shown). Thus, Gifsy-1 appeared to be fully competent for both lysogenic and lytic development. Subsequent improvement of the detection techniques using ‘green’ indicator plates (Maloy, 1990) also revealed that Gifsy-2 could be released as phage and form plaques. The data in Fig. 2 show that the two viruses are heteroimmune. However, the significantly smaller size of Gifsy-2 plaques on lawns of a Gifsy-1 lysogen (MA6275), compared with those formed on a doubly cured strain (MA5975; compare Fig. 2H and B), suggests that Gifsy-2 phage adsorption or proliferation is impaired in cells lysogenic for Gifsy-1. Similar ‘superinfection exclusion’ properties have been reported for other phages (Susskind et al., 1974; Groman and Rabin, 1980). A further indication of the interference exerted by Gifsy-1 on Gifsy-2 is the apparent failure of the latter to respond to induction in cells lysogenic for the former (Fig. 2G). This leads to a hierarchy in the order of induction of the two prophages, with Gifsy-1 being the only phage present in the supernatant from the wild-type strain (Fig. 2A and D). In contrast, Gifsy-2 is readily induced in cells that are cured for Gifsy-1, such as in strain MA5973 (Fig. 2B and H). Using the appropriate combinations of donor strains and non-immune recipient strains, we were able to isolate phage particles and purify Gifsy-1 and Gifsy-2 phage DNA for restriction enzyme analysis (Fig. 3). From these data, the sizes of the phage genomes could be estimated tentatively to be 50 kb and 45 kb for Gifsy-1 and Gifsy-2 respectively.

Figure 2.

. Detection of Gifsy-1 and Gifsy-2 viral particles and cross-immunity tests. Exponentially growing cultures were exposed to 2 mM H2O2 for 3 h and centrifuged. Supernatants were diluted 1:100 and spotted on bacterial cells spread on ‘green’ indicator plates. Phage infection foci appear as dark green specks on a light-coloured background. In the microphotograph, no specks are visible in (E), (G) and (I). In addition, no specks were ever seen with any of the above supernatants spotted on ATCC14028s or with supernatant from strain MA5975 spotted on any of the above strains (data not shown).

Figure 3.

. Restriction digestion and hybridization analysis of Gifsy-1 and Gifsy-2 phage DNA. Phage DNA was prepared as described in Experimental procedures and subjected to restriction nuclease digestion. A. Lanes 1 and 2, Pst I cleavage products of Gifsy-1 DNA (lane 1) and Gifsy-2 DNA (lane 2); lanes 3 and 4, hybridization of DNA fragments in lanes 1 and 2, respectively, to 32P-labelled DNA from plasmid pSY7 (sodC-specific probe). B. Lanes 1 and 2, HindIII cleavage products of Gifsy-1 DNA (lane 1) and Gifsy-2 DNA (lane 2); lane 3, hybridization of DNA fragments in lane 2 to the sodC-specific probe. C. Lanes 1 and 2, EcoRI cleavage products of Gifsy-1 DNA (lane 1) and Gifsy-2 DNA (lane 2); lane 3, hybridization of DNA fragments in lane 2 to the sodC-specific probe.

Gifsy-2-cured strains regain full virulence upon relysogenization

A patch of Gifsy-2-infected cells from a strain lysogenic for Gifsy-1 (MA6275) was resuspended and inoculated directly into mice (five animals each receiving ≈103 bacteria intraperitoneally). We determined afterwards that Gifsy-2-lysogenized cells represented between 1% and 2% of injected bacteria. Eight days later, all mice developed clear symptoms of terminal sepsis and were sacrificed. In all mice, 100% of the bacteria colonizing the spleens were Gifsy-2 lysogens and carried the prophage at its normal chromosomal position (see below). Thus, passage in a mouse offers an effective way of positively selecting such lysogens. The strain so obtained, MA6309, was re-inoculated into mice and found to have the same virulence properties as the original ATCC14028s parent (Table 2). These results confirm that Gifsy-2 contains virulence determinants and can mobilize them from one strain to another.

The Gifsy prophages are induced by hydrogen peroxide treatment

Wong and McClelland (1994) identified a sequence, designated RSP435, transiently transcribed during Salmonella adaptation to oxidative stress. The authors found this sequence to be duplicated in the chromosome and mapped the two copies at positions that closely correspond to the positions of the Gifsy prophages (Wong and McClelland, 1994). This prompted us to perform polymerase chain reaction (PCR) amplification experiments, which confirmed that the RSP435 locus lay within the prophages (data not shown). Suspecting that the activation of Gifsy sequences could reflect the process of prophage induction (Imlay and Linn, 1987), we tested whether hydrogen peroxide treatment elicited the release of phage particles. The results showed this to be the case. In our tests, hydrogen peroxide appeared to induce the Gifsy-1 prophage more effectively than mitomycin-C (at the concentration used for P22 prophage induction in S. typhimurium ; Youderian et al., 1988) or nalidixic acid (Table 3). Similar findings were made monitoring Gifsy-2 release from a Gifsy-1-cured strain (MA5973; data not shown). We determined subsequently that the RSP435 sequence corresponds to the C-terminal coding portion of the prophages' recE gene (see below).

Table 3. . Production of Gifsy-1 phage by S. typhimurium. a. Exponentially growing cultures of strain ATCC14028s were subjected to the indicated treatments for 3 h or left untreated.b. Aliquots (5 μl) from the chloroform-treated supernatants, straight or suitably diluted, were spotted on MA5973 bacteria spread on green plates. Phage titres were determined counting the number of dark specks (see Fig. 2) in the spotted area.Thumbnail image of

The Gifsy-2 prophage carries the sodC gene

While this work was in progress, two reports indicated a role for the sodC gene in Salmonella virulence (De Groote et al., 1997; Farrant et al., 1997). The product of this gene — periplasmic copper- and zinc-cofactored superoxide dismutase — catalyses the conversion of superoxide radicals to hydrogen peroxide, a potentially useful detoxification reaction in the bacterial defence against macrophage oxidative burst. The two studies included preliminary mapping of sodC and sequence data revealing the presence of phage tail-like genes near it (De Groote et al., 1997; Farrant et al., 1997). Examining these data, we realized that the sodC -spanning fragment included the BlnI restriction endonuclease recognition site at 23.8 cs of the genetic map, the only such site in the first one-third of the Salmonella chromosome. We have previously shown this site to be located within the Gifsy-2 DNA (Figueroa-Bossi et al., 1997). This suggested strongly that the sodC gene was carried by the prophage. This was confirmed by Southern analysis of Gifsy-2 phage DNA (Fig. 3) as well by PCR amplification experiments using chromosomal DNA from strains carrying or lacking this prophage (data not shown). More recently, a second sodC gene, named sodC-2, has been identified in Salmonella typhimurium (De Groote et al., 1998). Our results (not shown) indicated that sodC-2 is not present on either of the Gifsy prophages.

The sodC gene complements the lack of the Gifsy-2 prophage to an extent that depends on the presence of the Gifsy-1 prophage

Inactivation of the sodC gene has been shown to cause attenuation of Salmonella virulence. (De Groote et al., 1997; Farrant et al., 1997). These findings could be sufficient to explain the full contribution of the Gifsy-2 prophage to virulence. Accordingly, one might expect that the sodC gene alone should restore virulence in a strain cured for the Gifsy-2 prophage. To test this possibility, the sodC gene was amplified by PCR and integrated into the chromosome of strains lacking Gifsy-2 or both Gifsy-1 and Gifsy-2 prophages. For each strain, two constructs were made differing in the length of sodC flanking sequences and in the insert orientation relative to neighbouring chromosomal sequences (Δ[ilvI–leuO]3307 ::sodC and Δ[ilvI–leuO]3308 ::sodC ; see Experimental procedures for details). The resulting strains were inoculated separately into mice, orally in a initial test and later intraperitoneally. These studies showed all strains that had incorporated the sodC gene to have increased ability to persist in mouse spleens, causing a marked hypertrophy of these organs (three to four times larger than in the control animals). Intriguingly, however, the increase in virulence was much more conspicuous in strains that carried the Gifsy-1 prophage compared with strains cured for both prophages. This is evident in the results shown in Table 4 (compare strains MA6600 and MA6601 with strains MA6581 and MA6589 respectively). Altogether, these data indicate that the SodC protein accounts for part of the Gifsy-2 prophage contribution to Salmonella pathogenicity and strongly suggest that this prophage encodes additional virulence information that is duplicated on the Gifsy-1 genome.

Table 4. . Mouse virulence properties of Gifsy-2-cured strains carrying or lacking the sodC gene. a. All strains are derived from S. typhimurium strain ATCC14028s.b. Mice were inoculated intraperitoneally as described in Experimental procedures.Thumbnail image of

Further prophage features


A 13.3 kb DNA fragment covering one of the Gifsy-1 chromosomal junctions and a region of the prophage that contains sequences homologous to Gifsy-2 material was subjected to DNA sequence analysis. The sequence (Fig. 4A) reveals that the Gifsy-1 prophage is inserted within the coding sequence of the lepA gene, which specifies a GTP-binding protein of unknown function and was shown to be dispensable in Escherichia coli (Dibb and Wolfe, 1986). The lepA gene is not necessarily inactivated, as the prophage lies only 38 bp downstream from the initiation codon, and a functional gene might be regenerated upon insertion. The lepA reading frame is restored upon prophage excision (Fig. 4A). The Gifsy-1 sequence shows the overall organization of the lambdoid family. Adjacent to the attachment region lie putative int and xis genes with similarities to the corresponding loci of bacteriophage φ80. Interestingly, the analogies with φ80 extend to the organizational level. As for φ80, and unlike bacteriophages lambda and P22, the int gene of Gifsy-1 is in the opposite orientation of xis (Leong et al., 1986). Additional open reading frames (ORFs) show similarities to various phage or prophage genes, including a locus (Ehli1) reported to elicit a haemolytic phenotype in enteropathogenic E. coli (Stroeher et al., 1993). However, we have been unable to detect any specific haemolysis resulting from expression of the Gifsy-1 Ehli-1 fragment in S. typhimurium or E. coli.

Figure 4.

. A. Gene organization at the left end of the Gifsy-1 prophage (GenBank accession number AF001386). Putative genes (orfs) are indicated by arrows. Genes are oriented according to the left/right convention used in the prophage map of bacteriophage lambda (Szybalski et al., 1970). The sequence surrounding the chromosomal insertion site was determined on a PCR-amplified fragment from a strain cured for the prophage (MA4587). B. Genome organization of the pncB–pyrD interval in E. coli and S. typhimurium. The diagram compiles data from this work as well as from Frick et al. (1990), Blattner et al. (1997), De Groote et al. (1997) and from the Salmonella Genome Sequence Project, Genome Sequence Center, Washington University, St Louis, MO, USA.


The work of De Groote et al. (1997) indicated the presence, immediately adjacent to the sodC gene, of an ORF that, if translated, would specify a protein with strong homology to a family of bacterial outer membrane proteins implicated in mammalian cell adherence/invasion (ail ; Miller et al., 1990), intramacrophage survival (pagC ; Pulkkinen and Miller, 1991) and serum resistance (lom ; Barondess and Beckwith, 1990). This putative gene is therefore an additional candidate for a phage-borne virulence determinant.

Genetic mapping allowed us to position the Gifsy-2 prophage between the pncB and pyrD genes. Upon prophage curing, these two genes, which are normally unlinked by P22-mediated transduction in wild-type S. typhimurium, become about 60% co-transducible. In E. coli, the region between the pncB and pyrD genes contains the pepN gene (Blattner et al., 1997). Using Salmonella pepN sequence information obtained through the blast network service of the Genome Sequence Center, Washington University, St Louis, MO, USA, we performed PCR amplification experiments comparing DNA templates from strains carrying or lacking the Gifsy-2 prophage. This work allowed us to localize the prophage approximately in the middle of the 250 bp intercistronic region between divergently oriented pncB and pepN genes. Sequence analysis currently under way has identified the putative int gene adjacent to one of the chromosomal boundaries. Our preliminary data indicate that this gene bears no significant resemblance to the Gifsy-1 int gene and, unlike the latter, is oriented towards the outside of the prophage. Finally, this study highlighted some major organizational differences in the pncB–pyrD region between E. coli and S. typhimurium. In E. coli, the pncB and pepN genes are closely linked to each other, while the pepN and pyrD genes are separated by a 12 kb DNA segment. In S. typhimurium, pncB and pepN genes are separated by the 45 kb Gifsy-2 insert, while pepN and pyrD genes are contiguous (Fig. 4B).


The work described here provides evidence for bacteriophage involvement in S. typhimurium virulence. The two Gifsy phages appear to be fully functional and capable of transferring their pathogenicity-associated traits to susceptible bacteria. Both phages form tiny plaques, difficult to detect by standard assays. This, as well as the fact that most S. typhimurium isolates carry them as prophages and are therefore immune, might explain why they have gone unnoticed so far. At first glance, the two prophages differ greatly in their relative contributions to virulence. Curing bacteria for the Gifsy-2 prophage results in substantial attenuation, whereas removing Gifsy-1 has little effect on mouse virulence. Closer examination reveals that this difference reflects a lack of reciprocity in complementation: while Gifsy-2 can substitute fully for Gifsy-1, the opposite is not true. Most of this imbalance could be ascribed to a single gene present on the Gifsy-2 prophage and absent from the Gifsy-1 genome: the sodC gene. Providing the sodC gene alone was sufficient to make Gifsy-1 prophage contribution to virulence become apparent. Thus, these results suggest the presence on both prophages of additional virulence determinants that are duplicated or functionally interchangeable. We tested whether the ail-like locus flanking the sodC gene on the Gifsy-2 prophage could be one of these determinants. Southern analysis showed unambiguously that no ail-related sequence is present in the Gifsy-1 genome (data not shown). More work will be necessary to identify the Gifsy-1-encoded factors implicated in virulence.

The role played by the sodC gene in the Gifsy-2 contribution to Salmonella pathogenicity remains unclear. The product of this gene, copper- and zinc-cofactored superoxide dismutase, catalyses the conversion of superoxide radicals to hydrogen peroxide. Both the nature of this reaction and the periplasmic localization of the enzyme have suggested that SodC activity protects pathogenic bacteria from oxidative damage in host macrophages. Indeed, the identification of the sodC gene in S. typhimurium (Canvin et al., 1996) was closely followed by reports describing its contribution to virulence (De Groote et al., 1997; Farrant et al., 1997). Recently, a second sodC gene has been identified in S. typhimurium and designated sodC-2 (De Groote et al., 1998). This gene shows a high degree of sequence similarity to the unique sodC gene of E. coli (82% amino acid identities), unlike the original sodC (hereafter referred to as the Gifsy sodC ; 55% amino acid identities with the E. coli enzyme). These findings raise the question of the significance of having two different [Cu,Zn]-SodC enzymes in S. typhimurium. Perhaps expression or activity of each of the two proteins is optimized for different environments or conditions. Alternatively, the presence of the Gifsy SodC might be linked to some elusive aspect of bacteriophage biology. Although Gifsy-2 represents the first example of a SodC-encoding bacteriophage, the presence of sodC genes in viral genomes is not novel; such genes have been found in a number of animal and plant viruses, including baculovirus and vaccinia virus (Tomalski et al., 1991; Aguado et al., 1992). To date, the role of these genes remains unknown. In the system described here, an intriguing possibility arises from the observation that the two Gifsy prophages are particularly sensitive to induction by hydrogen peroxide. Hydrogen peroxide being the primary product of SodC enzyme activity, one might imagine that, under certain conditions, SodC activity can lead to prophage induction. This could occur in bacteria exposed to high levels of superoxide anions inside macrophages. The transcriptional burst associated with the induction process or the amplification of phage DNA during replication could increase the synthesis of phage-encoded virulence factors (Neely and Friedman, 1998; Waldor, 1998).

Experimental procedures

Bacterial strains

The strains used in this study are listed in Table 5. Wild-type S. typhimurium virulent strains were obtained from E. Groisman, Washington University, via A-B. Blanc-Potard (ATCC14028s) and from Françoise Norel, Pasteur Institute, Paris (SL1344 and C52). Bacteria were cultured as described previously (Figueroa-Bossi et al., 1997).

Table 5. . Bacterial strains (Yanisch-Perron et al. (1985); Palmer and Marinus (1994)). a. Where not specified, the Gifsy-1 or Gifsy-2 status of the Salmonella strain is [+].b. Where not specified, the source of the strain is this work.Thumbnail image of


Plasmid pK4114 is a pKK232-8 derivative carrying a 15 kb Pst I-generated DNA insert that includes ≈11 kb from the left end of the Gifsy-1 prophage (Fig. 4A; Figueroa-Bossi et al., 1997). Plasmid p6360 is a pUC18 derivative carrying a 3780 Pst I fragment from the S. typhimurium chromosome that includes most of the ilvI gene, the leuO gene and the leucine operon promoter (D. El Hanafi and L. Bossi, GenBank accession number AF117227). Additional plasmids were constructed in the course of this work: plasmid pSY1 was obtained upon ligating a PCR-amplified 814 bp DNA fragment that includes the sodC gene to the ‘T tails’ of vector pGEMT (Promega). Transferring the insert of plasmid pSY1 to vector pGEM11 (after cleavage of both plasmids with ApaI and NsiI enzymes) yielded plasmid pSY7. Plasmids pSY13, pSY14 and pSY15 were derived from plasmid p6360 as described below.

Genetic techniques

Transductional crosses using the high-frequency generalized transducing mutant of phage P22 (HT 105/1 int-201 ) were carried out as described previously (Maloy, 1990). Transformation of bacteria with plasmid DNA was accomplished by electroporation using a Bio-Rad Gene Pulser under the conditions specified by the manufacturer.

Gifsy prophage curing

The Gifsy prophage curing procedure took advantage of the properties of a particular mutation that destabilizes both prophages. Allele sbcE21 (a small deletion) disrupts a regulatory site within the Gifsy-1 prophage and results in the induction of an activity competent to excise both prophages (Figueroa-Bossi et al., 1997). A phage P22-transducing lysate prepared on strain MA4399 (Table 5) was used to transfer the sbcE21 deletion and linked Tn10dTc and MudF (Lac+) markers (also within the prophage) from the original LT2 background into the virulent strains. The lac segregation patterns of the resulting strains were monitored on Xgal indicator plates. Isolates that had lost Gifsy-1 were readily recognized from their Lac phenotype (white colonies). Isolates that had lost Gifsy-2 were identified from the fact that they formed solid blue colonies (unlike the blue/white-sectored colonies of the parental type) resulting from the lower frequency of Gifsy-1 excision in cells lacking Gifsy-2 (Figueroa-Bossi et al., 1997). Subsequent isolation of Lac derivatives from the latter class yielded strains cured for both prophages.

Preparation of Gifsy-1 and Gifsy-2 phage DNA

Patches of phage-infected bacteria (Fig. 2) were picked individually and spread onto LB plates. After overnight incubation at 37°C, plate contents were resuspended in LB, and bacterial cells were removed by centrifugation. Aliquots (0.5–2 ml) from the supernatants were added to 25 ml LB cultures of the appropriate recipient strains (OD600 ≈0.05), and the mixture was incubated at 37°C with vigorous shaking for several hours. Bacteria and cell debris were removed by centrifugation at 4°C for 30–45 min at 2000 × g. Supernatants were centrifuged at 35 000 × g for 90 min. Pellets were gently resuspended in 0.5 ml of saline phosphate buffer (pH 7.2) by shaking in the cold for several hours. Suspensions were centrifuged for 10 min at 7000 × g to remove residual bacteria. Supernatants were extracted once with chloroform, and aqueous phases were treated with DNAase I and RNAase A to final concentrations of 1 and 10 μg ml−1, respectively, for 30 min at 37°C. Samples were phenol extracted until the interface was clear. DNA was ethanol precipitated and resuspended in 100 μl of 10 mM Tris-HCl, 1 mM EDTA, pH 8.0.

Gel electrophoresis and Southern hybridization

Electrophoretic separation of DNA fragments in regular or pulsed-field agarose gels, DNA blotting and hybridization were carried out as described previously (Figueroa-Bossi et al., 1997).

PCR amplification and DNA sequencing

PCR amplification from bacterial colonies or from purified DNA was carried out using Taq DNA polymerase (Appligene-Oncor) according to standard protocols. The oligonucleotides used as primers for the PCR were synthesized chemically by Isoprim. DNA sequence analysis was carried out by Appligene-Oncor. Sequence comparisons were made using the blast algorithm (Altschul et al., 1997).

Virulence assays

Female BALB/c mice were inoculated when 7–8 weeks old. Bacterial strains were freshly streaked on LB agar (from frozen stocks) and their antigenic formula verified by slide agglutination using O- and H-specific rabbit antiserum (Pasteur Diagnostics). Bacterial cultures (in LB broth) were started from single colonies. After an overnight incubation at 37°C, cultures were diluted into fresh medium, and growth was resumed until an OD600 of 0.5. Bacterial cells were harvested and resuspended in PBS (pH 7.2) to a concentration of ≈1010 bacteria ml−1. Suitably diluted aliquots (0.2 ml) of this suspension were used for mouse inoculation. The number of injected bacteria was determined subsequently by plating on LB agar. Bacteria were delivered to mice by intraperitoneal injection or through an orally placed polyethylene catheter reaching the stomach of mildly anaesthetized animals. Mouse viability was recorded for at least 3 weeks. Spleens were removed and homogenized in 1 ml of sterile saline solution. Aliquots of the homogenate were diluted and plated on Drigalski-lactose or LB plates.

Amplification and cloning of the sodC gene

The sodC gene was amplified by PCR using Gifsy-2 phage DNA as the template. Two versions were produced differing in the length of flanking sequences. The ‘short version’ was obtained using oligonucleotides 5′-CCACCTTCGCGGACTGCTCC-3′ and 5′-CTGGAACCTAGGTGCTTGGC-3′ (De Groote et al., 1997). The amplified fragment (814 bp) included all of the sequence information necessary and sufficient to direct sodC gene expression in vivo (Farrant et al., 1997). This fragment was initially cloned into plasmid pGEM-T and subsequently transferred (as an ApaI–NsiI fragment) into vector pGEM11. The DNA insert was released by digestion with NsiI and BamHI restriction endonucleases, purified by agarose gel electrophoresis and ligated to plasmid p6360 DNA cleaved with Bcl I and NsiI enzymes. This led to the identification of plasmid pSY13, in which the segment between the Bcl I site and the distal NsiI site in the insert of plasmid p6360 is replaced by the sodC-spanning sequence. This work also identified plasmid pSY14, which results from the reinsertion of the NsiI–Bcl I fragment into the p6360 backbone and thus lacks the material between the NsiI sites. The ‘long version’ of the sodC gene was obtained with oligonucleotides 5′-GCTGTAAACCTGATGCATTACCACTGACGCGCCGACG-3′ and 5′-CCGTTTCAATTGGATATGCATACACAGATTTAAGCGGC-3′ as amplification primers. Both oligonucleotides differ from the respective complementary sequences by a single mismatch that introduces an NsiI enzyme recognition site. The amplified fragment (1485 bp) was cleaved with NsiI and Bgl II (a naturally occurring Bgl II recognition site is found 444 bp downstream from the sodC coding sequence). The resulting fragment (1378 bp) was ligated to p6360 DNA previously cut with NsiI and Bcl I. This led to the identification of plasmid pSY15, in which the sodC insert orientation is opposite to that in pSY13.

Chromosomal integration of the sodC gene

Integration of the sodC gene into the Salmonella chromosome relied on selecting recombination events between the sequences flanking the sodC insert in plasmids pSY13 and pSY15 and the corresponding sequences in the chromosome. Plasmids pSY13 and pSY15 were moved into strain MA6576. This strain lacks both Gifsy-1 and Gifsy-2 prophages and carries the leu operon promoter mutation leu-500 (Gemmill et al., 1984) flanked by two drug resistance markers: leu-3254 ::Tn10dTc, immediately downstream from the leu promoter; and leuO3256 ::Tn10dCm within the promoter-proximal region of the leuO gene. Because of the leu-500 and Tn10dTc markers, strain MA6576 is a leucine auxotroph. To select recombinants directly in plasmid-free cells, plasmid-harbouring cells were cultured for several generations and then lysed by phage P22. The phage lysate was used to transduce the same MA6576 strain or strain MA6594 (which lacks the Gifsy-2 prophage but contains Gifsy-1) selecting prototrophy. Leu+ transductants were obtained in all cases, and most of them were found to be TcS, CmS and ApS. The presence of the desired construction in some of these isolates was confirmed by PCR amplification tests. In the chromosome, the ‘short’ and ‘long’ versions of the sodC gene were designated Δ[ilvI-leuO]3307 ::sodC and Δ[ilvI-leuO]3308 ::sodC respectively. A similar procedure using plasmid pSY14 (above) and strain MA6475 as recipient allowed the construction of a strain in which the ilvI–leuO segment is deleted without being replaced by the sodC insert (Δ[ilvI-leuO]3306 ). The latter strain (MA6583) was useful for verifying that removal of the ilvI–leuO segment does not adversely affect Salmonella virulence.

Note added in proof

Sequence analysis of a DNA fragment from the right end of the Gifsy-2 prophage (GenBank accession number AF147699) shows a potential coding region for a protein 76% identical to Salmonella virulence protein MsgA (Gunn et al., 1995, J Bacteriol177: 5040–5047). Survey of the Salmonella typhimurium genome sequence database, Genome Sequence Center, Washington University, St. Louis, MO, USA, reveals that the Gifsy-1 prophage includes an ORF with sequence repeat motifs similar to Shigella invasion plasmid antigen gene ipaH (Hartman et al., 1990, J Bacteriol172: 1905–1915) and to the yopM gene of Yersinia (Boland et al., 1996, EMBO J15: 5191–5201).


We are indebted to Françoise Norel and Colette Coynault, Pasteur Institute, Paris, for advice and friendly encouragement throughout this work. We thank Stan Maloy, Dan Andersson and John Roth for comments on an early version of this manuscript. Stéphane Delmas and Yann Saint-George contributed critically to the cloning of the sodC gene and its introduction in the Salmonella chromosome. The expert technical assistance of Daniela Maloriol was much appreciated. This work was supported by the Centre National de la Recherche Scientifique (UPR 9061).