It is known that Yersinia pestis kills Caenorhabditis elegans by a biofilm‐dependent mechanism that is similar to the mechanism used by the pathogen to block food intake in the flea vector. Using Y. pestis KIM5, which lacks the genes that are required for biofilm formation, we show that Y. pestis can kill C. elegans by a biofilm‐independent mechanism that correlates with the accumulation of the pathogen in the intestine. We used this novel Y. pestis–C. elegans pathogenesis system to show that previously known and unknown virulence‐related genes are required for full virulence in C. elegans. Six Y. pestis mutants with insertions in genes that are not related to virulence before were isolated using C. elegans. One of the six mutants carried an insertion in a novel virulence gene and showed significantly reduced virulence in a mouse model of Y. pestis pathogenesis. Our results indicate that the Y. pestis–C. elegans pathogenesis system that is described here can be used to ide.jpgy and study previously uncharacterized Y. pestis gene products required for virulence in mammalian systems.
The soil nematode Caenorhabditis elegans has been used as a model host to study the pathogenic mechanisms of several human pathogens, including Gram‐positive and Gram‐negative bacteria, as well as fungal species (reviewed by Aballay & Ausubel, 2002; Ewbank, 2002; Alegado et al, 2003; Nicholas & Hodgkin, 2004b; Mylonakis & Aballay, 2005; Sifri et al, 2005). Recently, it was reported that Yersinia spp kill C. elegans by a mechanism that correlates with the formation of a biofilm in the head of the animals. It was proposed that biofilm production, which requires the hmsHFRS operon, prevents food intake and causes death by starvation (Darby et al, 2002). To ide.jpgy other virulence factors that are required for the biofilm‐mediated killing of Yersinia pseudotuberculosis, the virulence phenotype of 39 transposon insertion strains was studied in C. elegans. Seven mutants in unknown genes and in genes that encode factors involved in haemin storage and lipopolysaccharide biosynthesis were ide.jpgied (Joshua et al, 2003). A later report indicates that biofilm formation does not involve signalling between Yersinia and C. elegans and that it depends on a continuous exposure of the nematodes to Yersinia (Tan & Darby, 2004).
We sought to develop a Yersinia pestis–C. elegans pathogenesis model that aided ide.jpgication and characterization of virulence factors that are not necessarily related to biofilm formation but are important for mammalian pathogenesis. Using the Y. pestis strain KIM5, which lacks the hmsHFRS genes, and standard conditions for C. elegans infection, we show here that Y. pestis is able to cause a persistent and lethal intestinal infection in nematodes. We also show that several genes that were previously associated with virulence in other bacterial pathogens, ompT, y3857 and yapH, are required for the full virulence of Y. pestis in C. elegans. To ide.jpgy previously undescribed virulence factors, a Y. pestis EZ∷TN mutagenized library containing 984 clones was screened to ide.jpgy insertion mutants that show reduced virulence in nematodes. Two out of the six mutants isolated on the basis of their reduced virulence in C. elegans also showed reduced virulence in a mouse model of Y. pestis pathogenesis. One of these mutants harbours an insertion in a gene that has not been related to virulence before, the product of which is similar to a putative exported protein of Salmonella enterica serovar Typhi. These experiments show that C. elegans can be used as an alternative model host to accelerate the ide.jpgication and characterization of Y. pestis virulence factors that are required for pathogenesis in mammalian systems.
Results and Discussion
Y. pestis kills C. elegans in a biofilm‐independent manner
Y. pestis is the aetiological agent of bubonic plague, a zoonotic infection that is transmitted from natural animal reservoirs to humans, usually through the bite of an infected flea. In the infected flea, the digestive tract is blocked by a Y. pestis biofilm that is dependent on the same genes that are required for biofilm formation on the surface of C. elegans (Darby et al, 2002). We postulated that Y. pestis strains that are deficient in biofilm formation could enter the C. elegans intestine and eventually kill the animals, by a mechanism that is different from blocking food intake. As shown in Fig 1A, the nematodes died more quickly when fed on the commonly studied Y. pestis KIM5 strain than when fed on Escherichia coli OP50, the usual food source for growing C. elegans in the laboratory. Although Y. pestis KIM5 kills C. elegans more slowly than Y. pestis KIM6+ (Fig 1B), many other human pathogens, including S. enterica, Enterococcus feacalis, Serratia marcescens and Cryptococcus neoformans, also kill nematodes with similar kinetics (Aballay et al, 2000; Labrousse et al, 2000; Garsin et al, 2001; Mylonakis et al, 2002; Kurz et al, 2003). The time required for 50% of the nematodes to die (time to death 50; TD50) when fed on a Y. pestis KIM5 lawn at 25°C was calculated in five independent experiments and was determined to be 3.63±0.27 days, whereas the TD50 was 9.54±1.67 days when fed on an E. coli OP50 lawn. The TD50 values of C. elegans pathogens range from 4 h to 8 days, and almost half of the reported TD50 values are as high as 3.63, or higher (reviewed by Sifri et al, 2005). When eggs were placed on Y. pestis lawns, the animals reached adulthood, which indicated that Y. pestis, in spite of its virulence, supports nematode development and that nematode death is not related to any deficiency in food intake (data not shown).
To test whether the premature death of the nematodes that were infected with Y. pestis KIM5 correlates with the presence of bacteria in the intestine, we infected the animals with Y. pestis KIM5 expressing the Aequiorea victoria green fluorescent protein (GFP). Fig 1 shows that KIM5 is not capable of forming a biofilm on the surface of the animals and that the intestinal lumen of the nematodes is distended and full of bacteria 48 h after initial exposure to the pathogen. These results show that Y. pestis lacking the hmsHFRS genes is still virulent in C. elegans and that virulence factors other than those involved in biofilm formation can be studied using the Y. pestis–C. elegans pathogenesis system described here.
Characterization of the Y. pestis infection of C. elegans
Previous work has shown that certain human pathogens can resist the mechanical and humoral stresses of the C. elegans intestinal environment and persistently colonize the gut (Aballay et al, 2000; Labrousse et al, 2000; Garsin et al, 2001; Kurz et al, 2003). To determine whether Y. pestis KIM5 was capable of persistently colonizing the C. elegans intestine, nematodes were fed on Y. pestis KIM5/GFP for 24 h and then transferred to nonlabelled E. coli. Fig 2 shows the presence of Y. pestis/GFP in representative animals 24 h (panel A) or 48 h (panel B) after the animals were transferred to E. coli, whereas E. coli/GFP was not observed in a representative animal 24 h after the animal was transferred to E. coli (panel C). We also tested whether the persistent colonization of the C. elegans intestine by Y. pestis correlated with the killing of the animals. Using the transfer experiment described above, we showed that nematodes exposed to Y. pestis KIM5 for 24 h died at the same rate as when they were continuously fed on Y. pestis, whereas those transferred after an infection time of only 6 h were rescued from a lethal infection (Fig 2D). These results indicate that an initial infection period of 24 h is sufficient for Y. pestis to establish a persistent colonization in C. elegans.
The presence of a swollen tail is a morphological phenotype characteristic of nematodes that are infected with Y. pestis (Figs 1F, 2B). To our knowledge, the presence of a swollen tail has not been previously reported in C. elegans infected with other human pathogens, but it is very similar to that seen in animals that are infected with a specific C. elegans pathogen, Microbacterium nematophilum. These bacteria adhere to the anal region of the nematodes and induce localized swelling of the underlying hypodermal tissue (Hodgkin et al, 2000). After initial characterization of this infection, it was found that the extracellular signal‐regulated kinase–mitogen‐activated protein kinase (ERK–MAPK) signalling pathway mediates both tail swelling and a protective response against M. nematophilum attack (Nicholas & Hodgkin, 2004a). It will be interesting to use the large collection of mutants in the ERK–MAPK signalling pathway to determine whether a similar C. elegans defence response is required to fight Y. pestis infection.
Known Y. pestis virulence factors in the killing of C. elegans
As many Y. pestis virulence determinants are encoded on virulence plasmids, we decided to study the role of plasmids pCD1 and pPCP1 on Y. pestis virulence in C. elegans. The pCD1 plasmid (also called pYV) encodes a type III secretion system (TTSS) required for full virulence in mammalian systems, although it is not sufficient to cause disease (Heesemann & Laufs, 1983; Heesemann et al, 1984). The pPCP1 plasmid encodes the plasminogen activator protease, which is important for dissemination of Y. pestis after subcutaneous injection into mammalian hosts (Wren, 2003). To determine the importance of these plasmids in Y. pestis virulence in C. elegans, we compared the susceptibility of the nematodes to strains cured of one or both plasmids. Fig 3A shows that strains lacking one or two virulence plasmids showed virulence that was comparable to that of Y. pestis KIM5, which carries both virulence plasmids. These results show that virulence factors encoded in pCD1 and pPCP1 are not required for Y. pestis virulence in C. elegans.
In addition to plasmid‐encoded virulence factors, several chromosomally encoded virulence factors have been ide.jpgied in Y. pestis, including genes encoding a potential second TTSS (Revell & Miller, 2001). As shown in Fig 3B, at least three Y. pestis KIM5 deletion mutants in chromosomally encoded genes presumably related to virulence (ompT, y3857 and yapH) that were available in our stock showed reduced virulence in nematodes. ompT encodes a member of the omptin family of outer‐membrane proteases, which have been implicated in the virulence of E. coli, S. enterica, Shigella flexneri and Y. pestis virulence (Kukkonen & Korhonen, 2004). y3857 encodes a putative exported metalloprotease that shows significant amino‐acid sequence similarity to virulence‐related metalloproteases from Yersinia ruckeri, Photorhabdus luminescens, S. marcescens and Proteus mirabilis (Walker et al, 1999; Fernandez et al, 2002; Bowen et al, 2003). The yapH gene encodes a putative autotransporter adhesion protein; the importance of this family of proteins in bacterial attachment to the surface of host cells and protection from complement and antimicrobial peptides has been studied in several species (Roggenkamp et al, 2003; Rose et al, 2003; Wehrl et al, 2004).
Ide.jpgication of novel mammalian virulence factors
To ide.jpgy novel Y. pestis virulence factors, an EZ∷TN insertion library containing 984 Y. pestis clones was screened, as described previously (Tenor et al, 2004), for mutants that failed to kill or showed attenuated killing of C. elegans. Six out of 984 insertion mutants were ide.jpgied on the basis of their slower killing of C. elegans than the killing by wild‐type Y. pestis KIM5. The molecular characterization of the isolated strains indicates that all the mutations are in genes that have previously not been implicated in Y. pestis virulence. Three of the insertions, however, are in genes that are related to virulence in other species (Table I).
To determine whether Y. pestis virulence‐related genes ide.jpgied using C. elegans as a host have a role in mammalian pathogenesis, we used an intranasal mouse model of Y. pestis pathogenesis (supplementary information online) to study the virulence phenotypes of four isolated mutants. Using Y. pestis KIM5, we found that the medial lethal dose (MLD or LD50) is about 105 CFU. After intranasal inoculation of 10 LD50, bacteria are present in the lungs at relatively low numbers for several days, but the lung infection is often cleared by days 4–6. Histology on days 3 and 4 after intranasal inoculation shows micro‐abscesses in the spleen and liver. These abscesses enlarge, leading to the complete destruction of the spleen architecture in some mice by days 6–8 (Fig 4). Although the route of infection is the lungs, this model is more representative of septicemic than of human pneumonic plague. Fig 4 shows the virulence phenotypes of two mutants with reduced virulence both in the C. elegans model and in the mouse model of Y. pestis pathogenesis. The virulence phenotype of Y. pestis 9F11 is provided as an example of a mutant in a gene that is probably required for full virulence in C. elegans but not in mammals. The reduced virulence of Y. pestis KIM10 indicates that the intranasal inoculation of bacteria alone does not cause lethality.
In summary, two out of the four mutants that were studied in the mouse model showed a statistically significant attenuation in virulence. The attenuated virulence shown by Y. pestis 5G5 is not surprising, as y1021 encodes a cytochrome subunit previously implicated in S. enterica virulence (Turner et al, 2003). The reduced virulence of Y. pestis 5B9 is interesting because y0340 has not previously been related to virulence and its product is a member of an uncharacterized family of putative exported proteins that are found exclusively in several pathogens in the family Enterobacteriaceae. Y. pestis KIM5 encodes seven individual orthologues of the y0340 gene product in its chromosome. In fact, y0340 is the last gene in an operon encoding three of the seven Y0340 orthologues. The gene downstream of y0340 is predicted to be transcribed in the opposite direction of y0340, making the possibility of a polar effect due to the mutation highly unlikely. These results show that C. elegans can be used as an alternative model host to accelerate the ide.jpgication and characterization of Y. pestis virulence factors required for pathogenesis in mammalian systems.
One drawback of using C. elegans to model bacterial pathogenesis is the fact that the nematode is grown at temperatures below 25°C. It is well established that exposure to temperatures of around 37°C in the mammalian hosts results in the upregulation of a range of Y. pestis virulence factors. In contrast, other Yersinia virulence factors, such as invasin, and the Y. enterocolitica chromosomal YSA TTSS are preferentially expressed at 26°C (Revell & Miller, 2001). In addition, Y. pestis virulence factors that are required for colonization of the flea are optimally expressed at lower temperatures. For example, the plasmid pMT1‐encoded Ymt protein, which has phospholipase D activity, is preferentially expressed at 26°C and has an important role in enabling infection of the flea (Hinnebusch et al, 2002).
Soon after being injected by the flea, Y. pestis has to deal with mammalian defences; therefore, the bacterium probably has to express proteins that are required for defence against many mammalian innate defences at 25°C. For example, Y. pestis is constitutively serum resistant, whereas, in the enteropathogenic Yersiniae, serum resistance is expressed only at 37°C. This could also be required owing the presence of serum‐like products in the flea gut. Thus, although temperature is a limitation, we theorized that C. elegans could be useful for studying virulence factors that are optimally expressed at 25°C or constitutively expressed at both 25 and 37°C. In spite of the long evolutionary distance and differences in growth temperatures between nematodes and mammals, the results reported here show that C. elegans can be used as an alternative host to study Y. pestis virulence factors that are required for pathogenesis in mammalian systems. These types of virulence factor expressed at low temperatures may be required to regulate innate immunity pathways conserved in nematodes, fleas and mammals.
Bacterial strains and growth conditions. E. coli OP50 (Brenner, 1974), Y. pestis pgm− strains KIM5, KIM6, KIM8 and KIM10 (Une & Brubaker, 1984) and Y. pestis pgm+ strain KIM6+ (Perry et al, 1990) have been described previously. Y. pestis KIM5ΔompT, Y. pestis KIM5ΔyapH and Y. pestis KIM5Δy3857 were generated using the phage lambda Red recombinase method (Datsenko & Wanner, 2000), essentially as described previously (Jackson et al, 2004).
Statistical analyses. Animal survival was plotted using the PRISM (version 4.00) computer program. Survival curves are considered significantly different from the control when P‐values are <0.05. Prism uses the product limit or the Kaplan–Meier method to calculate survival fractions and the log‐rank test, which is equivalent to the Mantel–Heanszel test, to compare survival curves. Supplementary information is available at EMBO reports online (http://www.nature.com/embor/journal/vaop/ncurrent/extref/7400516‐s1.pdf).
We thank E. Kintz for technical assistance. G.P. is supported by National Institutes of Health (NIH) Grants RO1AI39575 and RO1AI50552. A.A. is supported by the Whitehead Scholar Program, the Duke Center for Translational Research, the NIH Southeast Regional Center of Excellence for Biodefense & Emerging Infections, and NIH Grant RO1GM070977.
- Copyright © 2005 European Molecular Biology Organization