Phage therapy has been used in the past as an alternative therapy against bacterial pathogens. However, phage-resistant bacterial strains can emerge. Some studies show that these phage-resistant strains are avirulent. In this study, we report that phage-resistant strains of Salmonella enterica serovar Enteritidis (hereafter S. Enteritidis) were avirulent in the Caenorhabditis elegans animal model. We isolated phage-resistant strains of S. Enteritidis ATCC 13076 by using three lytic phages (f2αSE, f3αSE and f18αSE). In these mutants, we explored different virulence factors like lipopolysaccharide (LPS), virulence plasmid (Pla), motility and type I fimbriae, all of which may have effects on virulence and could furthermore be related to phage resistance. The phage-resistant strains of S. Enteritidis showed loss of O-Polysaccharide (O-PS) and auto-agglutination, present a rough phenotype and consequently they are avirulent in the C. elegans animal model. We speculate that the O-PS is necessary for phage attachment to the S. Enteritidis cell surface.

Salmonella enterica serovar Enteritidis (hereafter S. Enteritidis) causes gastrointestinal disease in humans and is a major public health concern due to its ability to be transmitted via contaminated eggs or egg based and poultry meat products (Dominguez et al. 2002). An approach to the control of S. Enteritidis is the use of bacteriophages, which have proven of value in the curtailment of S. Enteritidis infection in Cheddar cheese (Modi et al. 2001), vegetables (Leverentz et al. 2001), poultry products (Higgins et al. 2005) and the skin of chickens (Goode et al. 2003). In this latter context, we have recently isolated and described three dsDNA phages that lyse S. Enteritidis in vitro and additional variants of these phages that lyse S. Pullorum (Santander and Robeson, 2002). These phages have a morphology similar to bacteriophage λ of Escherichia coli. Furthermore, they form clear plaques in lawns of the Salmonella serovars mentioned above (Santander and Robeson, 2002). In addition, we tested these phages in phage prophylaxis assays in C. elegans (Santander and Robeson, 2004). All phages tested (f2αSE, f3αSE and f18αSE) protect C. elegans from infection and subsequent death by S. Enteritidis. We also reported that S. Pullorum was able to infect and kill C. elegans and that f3αSP, a variant of f3αSE adapted to S. enterica serovar Pullorum, protects C. elegans from S. enterica serovar Pullorum killing (Santander and Robeson, 2004). Furthermore, recent studies showed that bacteriophage f3αSE persists in the avian system (Krüger et al. 2003) and reduces the colonization by S. Enteritidis in chicks (Borie et al. 2004) in a similar vein to the report of Fiorentin et al. (2005).

However, phage resistant strains could emerge and persist. Studies on phage therapy against Pseudomonas pecoglossida, a fish pathogen, showed that the phage-resistant bacterial strains lost their virulence (Park et al. 2000). Based on this observation we isolated phage-resistant strains of S. Enteritidis and used them to challenge C. elegans. We determined that the phage-resistant strains of S. Enteritidis were avirulent in the C. elegans model and therefore reasoned that phage resistance and virulence towards C. elegans should be related. Other bacteriophages of Salmonella, as P22, have their attachment site at the lypopolysaccharide (LPS) (Steinbacher et al. 1997). Furthermore, strains of S. Enteritidis such as PT7 do not express the long-chain LPS, show auto-agglutination and they are less virulent than S. Enteritidis PT4 in Balb/C mice (Chart et al. 1989). Thus, the loss of virulence in our S. Enteritidis phage-resistant strains could be due to a change in the LPS molecule. We determined that the strains resistant to f2αSE, f3αSE and f18αSE lost the O-Polysaccharide (O-PS) and show auto-agglutination; consequently, they are unable to kill C. elegans. We assume that the O-PS is involved in phage attachment during early infection by these phages in S. Enteritidis.

Salmonella strains, bacteriophages and conditions of culture

Salmonella strains used in this study are listed in Table 1. Strains were routinely cultured at 37ºC in LB medium (Bacto Tryptone, 10 g/liter; Bacto Yeast extract, 5 g/liter; NaCl 5 g/liter) (Sambrook and Russell, 2001) or Nutrient broth (Difco). Media was solidified with 1.5% (wt/vol) agar. When required, the medium was supplemented with streptomycin (Sm; 25 µg/ml), mannose (mann 0.5% wt/vol) or galactose (gal 1.0% wt/vol). Phages used in this study are listed in Table 2. Liquid lysates in LB broth (10⁹ - 10¹⁰ pfu/ml) were propagated in S. Enteritidis ATCC 13076 as a host, using pump-aerated cultures at 37ºC (Santander and Robeson, 2002). The final phage suspension was treated with chlorophorm (5 µl/ml), titrated and kept at 4ºC.

Isolation of phage-resistant strains of S. Enteritidis

An exponential culture of S. Enteritidis (2 x 10⁸ cfu/ml) was infected with individual strains of phages and a mixture of them. All the infections were made at a multiplicity of infection (MOI) of 1. The mixtures were incubated at 37ºC for 1 hr and plated onto LB agar. The resistant colonies obtained were reisolated. Green indicator plates were used to confirm that the phage resistant strains were phage-free (Provence and Curtiss, 1994). The resistant strains were tested for phage resistance stability and susceptibility to the other phages.

Characterization of the phage-resistant strains of S. Enteritidis

The S. Enteritidis strains were characterized for type I fimbriae in static broth cultures (Leathart and Gally, 1998) and motility in motility medium (bioMérieux, Marcy I'Etoile, France). LPS presence was evaluated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and visualized by silver staining (Hitchcock and Brown, 1983). The presence of O-PS was corroborated by Western blot analysis (Harlow and Lane, 1999) of D₁ (O_1,9,12) using rabbit D₁ polyclonal (1:5000) (Difco) as primary antibody, and alkaline phosphatase-conjugated anti-rabbit IgG (1:10000) (Sigma) as secondary antibody. Plasmid profiles were verified by alkaline lysis and agarose gel (0.5%) electrophoresis (Sambrook and Russell, 2001). Fermentation patterns of various carbohydrates and production of H₂S were determined by using the API 20E system (bioMérieux, Marcy I'Etoile, France). The S. Enteritidis strains gave API 20E code number 670455257 (good identification as Salmonella spp. 89.0%). Phage typing characterization was done by the macroplaque assay using the ten traditional typing phages (Ward et al. 1987). Agglutination tests were performed on glass microscope slides by mixing 50 µl of antisera against D₁ (O₁,₉,₁₂) and H_m (Difco Laboratories, Detroit, MI) with suspensions of fresh single colonies. Reactions were visualized by phase-contrast microscopy at 10X magnification.

C. elegans maintenance

The nematode C. elegans Bristol N2 was maintained in modified NG agar (0.35% bacto peptone) and fed with 200 µl of a fresh culture of Escherichia coli OP50 per plate (Hope, 1999).

We used the method described by Aballay et al. (2000), without transfers (Santander and Robeson, 2004). Nematodes were incubated at 25ºC for 10 days. Dead nematode counts were performed every 24 hrs eliminating dead specimens from the plate. Thus, we determined the TD₅₀ (Time it takes for 50% of the nematodes to die).

Recently we determined that S. Enteritidis and S. Pullorum kill C. elegans (Santander and Robeson, 2004). Furthermore, we used the C. elegans model to evaluate protection due to phage prophylaxis against these Salmonella strains (Santander and Robeson, 2004), using three lytic phages isolated and characterized before (Santander and Robeson, 2002).

Studies with P. pecoglossida had shown that phage resistant strains become avirulent for their fish host (Park et al. 2000), suggesting that the phage attachment site at the bacterial cell surface could be related to pathogenicity. We thought a similar situation could apply to S. Enteritidis. To test this hypothesis we isolated S. Enteritidis strains resistant to f2αSE, f3αSE and f18αSE and to a mixture of all three phages. We found that all S. Enteritidis phage-resistant strains that we isolated were phage-free. These strains were stably resistant to the corresponding phage and were also resistant to the other phages. These results suggested that the attachment site at the surface of S. Enteritidis is the same for the three phages assayed. In addition, the phage resistant strains were phage-typed with the ten traditional phages used for phage-typing S. Enteritidis (Ward et al. 1987) and were shown to be resistant to all the phages utilized in the phage-typing test (Table 3), suggesting a common attachment mechanism for all these typing phages as well. Furthermore, S. Enteritidis PT7, a derivative of S. Enteritidis PT4 which is unable to express the long-chain LPS (Chart et al. 1989), is resistant to f2αSE, f3αSE and f18αSE (Santander and Robeson, 2002). However, S. Enteritidis PT7 has a different phage-type profile in comparison with our S. Enteritidis 13076 phage-resistant mutants (Table 3), perhaps because some of the traditional typing phages do not interact with the O-PS as f2αSE, f3αSE and f18αSE apparently do during phage infection.

In the context of phenotypic variation in Salmonella we were aware that conversions in phenotype are associated to acquisition or loss of either a temperate phage or a plasmid in S. Enteritidis (Baggesen et al. 1997; Gregorova et al. 2002). Therefore, we thought that phage-resistance in our S. Enteritidis 13076 isolates could be connected to such phenomena. However, our phage resistant strains preserve their plasmid content (Figure 1) and were tested to be phage-free using green agar plates. These results indicate that our phage-resistant derivatives are mutants obtained by phage-mediated selective pressure against the wild type strain with a complete O-PS.

Infection experiments with C. elegans showed that the S. Enteritidis phage-resistant strains were unable to kill the worm in contrast to the wild-type strain (Figure 2). Both a f3αSE resistant strain and a strain isolated as a triple phage-resistant mutant, were avirulent in the C. elegans assay (Figure 2). Since our data indicated that phage resistance and loss of virulence were related, we characterized the main virulence factors of phage-resistant strains, which were likely to change. All the phage resistant strains presented phenotypic markers similar to the wild-type except that the phage-resistant strains showed auto-agglutination, which made impossible the detection of somatic and flagellar antigens (Table 1). However, these strains were non-motile and have the virulence plasmid (Pla), an important virulence factor in Salmonella (Rychlik et al. 2006; Figure 1). The auto-agglutination test indicated that the LPS had changed. The LPS profiles showed a notable difference between the phage-resistant strains and the wild-type, as expected (Figure 3a). Therefore, the phage-resistant strains of S. Enteritidis 13076 presented a rough phenotype. Western blot analysis also showed that the S. Enteritidis phage resistant strains had lost the O-PS (Figure 3b). These results indicated that motility may not be required by S. Enteritidis to kill the worm and that the O-PS is an important virulence factor for the infection and subsequent killing of C. elegans.

It is known that S. Typhimurium requires an intact O-PS to trigger programmed cell death in C. elegans and consequently kill the worm (Aballay et al. 2001; Aballay et al. 2003). This last study is in agreement with our observations that the S. Enteritidis phage-resistant mutants with an incomplete O-PS (Figure 3) fail to kill C. elegans (Figure 2).

We then sought to investigate further the nature of the genetic defect that led to the rough phenotype of the S. Enteritidis phage-resistant mutants. It is known that manE (or pmi) Salmonella mutants have a defect in phospho-mannose isomerase which cannot convert, mannose into mannose-6P and consequently, in the absence of mannose their O-PS is incomplete (Collins et al. 1991). Also, there are mutants in the galE gene encoding UDP-galactose-4-epimerase which converts UDP-glucose in UDP-galactose, leading to the synthesis of the LPS O-antigen side chain and core. Therefore, in the absence of galactose, galE mutants become rough. However, addition of small amounts of galactose to the growth medium of these mutants results in production of sufficient UDP galactose to restore LPS production; but if galE mutants are fed excess galactose they accumulate UDP galactose, which is toxic to the cell (Adhya, 1996). To test whether any of these phenotypic changes had occurred in the S. Enteritidis phage-resistant mutants, we grew these strains in nutrient broth, which in contrast to LB broth does not have galactose or mannose (data not shown). We found that the strains grew in the presence and absence of both mannose and galactose and that these sugars did not have any effect on the LPS profile (data no shown). These facts indicate that defects in the galE (or the galK or galU related genes) and the manE gene are not the cause of the incomplete O-LPS chain.

These results collectively imply that S. Enteritidis avirulence in C. elegans and phage-resistance are related. Consequently, in connection with phage therapy or prophylaxis, use of the bacterial viruses referred to in this study should not be hampered by the eventual emergence of phage-resistant mutants, since they become avirulent. However, should they remain a sanitary problem it is always possible to search for additional phages to which they may serve as hosts and wherein such phages may presumably target outer membrane proteins.

We are particularly grateful to Dr. Roy Curtiss III (Arizona State University, Biodesign Institute) for providing his lab facilities to perform some of our experiments and Dr. Melha Mellata (Arizona State University, Biodesign Institute) for her criticism and assistance in determining large plasmid profiles.