Identification and Characterization of Cri1, a Locus Controlling Mortality During Citrobacter Rodentium Infection in Mice

ORIGINAL ARTICLE Identification and characterization of Cri1, a locus controlling mortality during Citrobacter rodentium infection in mice

E Diez1,2, L Zhu1,2, SA Teatero1,2, M Paquet3, M-F Roy4, JC Loredo-Osti5, D Malo2,6 and S Gruenheid1,2 1Department of Microbiology and Immunology, McGill University, Montreal, Quebec, Canada; 2Complex Traits Program, McGill University, Montreal, Quebec, Canada; 3Veterinary Comparative Pathology Services, Comparative Medicine and Animal Resources Centre, McGill University, Montreal, Quebec, Canada; 4Department of Veterinary Clinical and Diagnostic Sciences, Faculty of Veterinary Medicine, University of Calgary, Calgary, Alberta, Canada; 5Department of Mathematics and Statistics, Memorial University, St John’s, Newfoundland and Labrador, Canada and 6Departments of Medicine and Human Genetics, McGill University, Montreal, Quebec, Canada

Infection of inbred mouse strains with Citrobacter rodentium represents an ideal model to reveal the genetic factors controlling host resistance to noninvasive enteric bacterial pathogens. We have chosen a positional cloning approach to identify putative gene(s) that control the known difference in survival between resistant C57BL/6J and susceptible C3H/HeJ and C3H/HeOuJ mice. Our work has identified one major locus within proximal chromosome 15 that is responsible for the marked susceptibility of both C3H strains, and we formally exclude Tlr4 from control of survival to this pathogen. We have named this new host resistance locus Cri1 (Citrobacter rodentium infection 1). The Cri1 genetic interval currently spans B16 Mb and it confers survival to the infection in a recessive manner. Transfer of the Cri1 locus from the surviving B6 mice into a congenic mouse with a C3Ou genetic background confirms its overall chromosomal localization and its highly significant effect on host survival. The C3Ou.B6-Cri1 mice thus produced have also enabled us to dissociate the control of mouse survival from the control of bacterial load early in the infection as well as from control of colonic hyperplasia. Genes and Immunity (2011) 12, 280–290; doi:10.1038/gene.2010.76; published online 17 February 2011

Keywords: QTL; Citrobacter rodentium; mouse; survival; Cri1

Introduction increase.10 These observations are all the more worri- some as there are currently no specific therapeutic Citrobacter rodentium infection of mice was first described regimens for EHEC infections; antibiotic use is not in the 1960s and 1970s as the suspected causative agent recommended as it exacerbates disease and increases of outbreaks of diarrhea, colitis and colonic hyperplasia the rate of development of complications.11 in mouse colonies in the United States and Japan.1–5 More The parallels between C. rodentium infection of mice recently, infection of mice with this intestinal pathogen and EHEC and EPEC infection of humans are striking. has become recognized as an excellent model for These Enterobacteriaceae colonize the gastrointestinal infection with the human-specific diarrheal pathogens tract of their hosts by the formation of characteristic enterohemorrhagic and enteropathogenic Escherichia coli attaching/effacing lesions, leading to the development of (EHEC and EPEC; reviewed in Mundy et al.6). EPEC kills diarrhea, colitis and hyperplasia of the intestinal epithe- several hundred thousand children each year in devel- lium. Using a type III secretion system that is conserved oping countries,7 whereas EHEC is commonly associated between all three bacterial species, they inject bacterial with foodborne diarrheal outbreaks in the developed virulence proteins directly into host cells. Recent geno- world.8 Over 110 000 cases of EHEC infection and 61 mic and proteomic characterization of the secreted EHEC-associated deaths are estimated to occur each year protein repertoires in EHEC, EPEC and C. rodentium in the United States alone.9 Furthermore, recent epide- indicates that the vast majority of injected virulence miological data indicate that both the frequency of EHEC proteins are common to all three bacterial species.12,13 outbreaks and the severity of disease suffered are on the Consequently, it is not surprising that the C. rodentium model is now widely used to investigate the role of bacterial virulence proteins and host defense mechan- Correspondence: Dr S Gruenheid, Department of Microbiology and isms in the attaching/effacing pathogenesis. Immunology, McGill University, 3649 Promenade Sir William Osler, Studies in humans indicate that the disease severity Room 365, Montreal, Quebec, Canada H3G 0B1. E-mail: [email protected] and clinical outcome of EHEC and EPEC infection can Received 16 March 2010; revised 14 July 2010; accepted 24 August vary greatly between individuals.14–17 Whether these 2010; published online 17 February 2011 differences have a genetic basis is undetermined. On the Cri1 and susceptibility to C. rodentium infection E Diez et al 281 other hand, differences in susceptibility to C. rodentium 100 infection in mice were first described in 1977, when an abnormally high mortality rate was noted in the 80 C3H/HeJ strain.18 Later studies revealed that C3H/HeJ mice are highly susceptible to Salmonella infection and are hyporesponsive to bacterial lipopolysaccharide be- 60 cause of a mutation in Toll-like receptor 4 (Tlr4), raising the question of whether the Tlr4 defect was also responsible 40 for C. rodentium susceptibility. However, the finding Percent survival that the closely related but Tlr4-wild-type mouse strain B6 C3H/HeOuJ is also susceptible to C. rodentium infection 20 C3Ou suggested that C. rodentium susceptibility was indepen- C3 19 dent of Tlr4. Additional support for this was provided 0 by side-by-side phenotypic studies of Tlr4-deficient and 0 5 10 15 20 25 30 20 Tlr4-wild-type mice. In this study we apply a systema- Days post infection tic genetic approach to the C. rodentium infection model to investigate the genetic basis for high mortality in B6 10 C3H/HeJ and C3H/HeOuJ mice. C3Ou C3 8 Results 6 Phenotypic differences between C57BL/6J and C3H inbred mice during C. rodentium infection 4 The inbred mouse strains C3H and C57BL/6J (B6) differ in their susceptibility to C. rodentium infection.18,19 The 2 most striking difference being that mice of the C3H/HeJ (C3) and C3H/HeOuJ (C3Ou) strains succumb to the infection within 12 days following orogastric infection, Bacterial shedding (log cfu / g of feces) 0 0 5 10 15 20 whereas B6 mice not only survive but fully recover and clear the infection (cumulative survival curves, Days post infection Figure 1a). Besides the obvious difference in survival Figure 1 Phenotyping of inbred mouse strains following Citrobac- between the B6 and both C3H strains, the C3Ou strain ter rodentium infection. (a) Cumulative survival in the resistant displays a reproducible and statistically significant C57BL/6J and the susceptible C3H/HeJ (C3) and C3H/HeOuJ (C3Ou) inbred mouse strains following gastric gavage with 2.5 Â 108 increase in survival time when compared with the C3 CFUs of C. rodentium. A total of 10 animals from each strain (5 males strain (Po0.0001 with a log-rank statistical test). Com- and 5 females) were infected. No significant differences in survival pared with B6 mice, C3 and C3Ou mice also display a were noted between the sexes. Besides the obvious difference in much more rapid rate of colonization of the colon by survival between the C57BL/6J and both C3H strains, the C3Ou C. rodentium, leading to significantly greater bacterial strain displays a reproducible and statistically significant increase in loads early on in infection.19 Figure 1b shows a time survival time when compared with the C3 strain (Po0.0001 log- rank statistics). (b) Bacterial shedding in fecal pellets from infected course of C. rodentium shedding in the surviving B6 inbred mice. Bacterial shedding is measured as C. rodentium CFU g–1 mouse strain and in the susceptible C3 and C3Ou strains. of homogenized fecal matter. Five mice from each inbred strain Two important observations can be made from the data were infected and transferred into bedding-free isolation cages at obtained. First, the colony-forming unit (CFU) counts the specified time points to collect six fresh fecal pellets from each. increase significantly faster in both C3H strains com- Bacterial shedding was measured every 24 h after infection. Data pared with B6 mice (Po0.05 at days 3–5 post infection), collection was stopped for the C3H strains at day 7, when signs of 19 morbidity/mortality appeared. For the B6 mice, CFU counts were and that is consistent with previously published data. further compiled at days 10, 14 and 21 post infection to show Second, B6 mice reach similar high levels of C. rodentium bacterial clearance, as described previously.33 Mean log-transformed shedding in fecal pellets before they start clearing the bacterial counts are shown in the graph and error bars depict s.d. infection after day 10 post infection. Although it has been Nonlinear curve fitting was performed using fourth-order hypothesized that the greater bacterial loads early in polynomial functions. infection may lead to mortality in C3H strains,19 the link between these two phenotypic differences remains an open question. pellets. Colonic hyperplasia and hemorrhage are evident In order to gain more insight into the phenotypes as well, even at the macroscopic level. In contrast, B6 under study, the parental mouse strains were examined tissues display low levels of hyperplasia and only in detail at day 7 post infection. This is the time point slightly increased water content in their fecal pellets. closest to the peak of infection in C57BL/6J mice where Although B6 mice are undeniably sick at that point, it is there are still some surviving C3 mice (Figure 1a). Mice nowhere near the phenotype observed in C3H mice. We from each of the infected inbred strains were killed and hypothesized that the observed phenotypic differences their cecum and colon dissected. At a macroscopic level between the mouse strains may be controlled by discrete (representative photographs in Supplementary Figure 1), host genetic factors. In order to gain insight into the tissues from both C3H mouse strains display obvious factors controlling these phenotypic differences, and to symptoms of disease: the cecums are highly retracted investigate the links between them, we set out to identify and the colons are devoid of properly formed fecal the genetic loci controlling the survival phenotype of B6

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 282 mice versus both the C3 and C3Ou strains. We then F1 hybrid mice were obtained by crossing resistant B6 determined if the loci controlling survival also affect the and susceptible C3 mice. These F1 mice were infected other phenotypic differences observed. and survival was recorded: the cumulative survival curve obtained was intermediate relative to that of the parental strains, with 55% survival at day 30 post Segregation of survival phenotype in F2 crosses infection (Figure 2a). Very little mortality occurs beyond We undertook a systematic positional cloning effort by day 20 post infection throughout our studies. We are crossing the resistant B6 mice with either of the therefore comfortable with considering that mice that susceptible C3 and C3Ou strains and correlating inheri- survived to day 30 are on the way to clearing the infection tance patterns throughout the genome with the resulting and are indeed ‘survivors’. A total of 168 progeny from a survival phenotypes. Data analysis for the crosses to each (C3 Â B6)F2 cross (hereafter named C3B6F2) were simi- of the two C3H mouse strains was kept separate larly infected and survival was recorded. The F2 throughout this study for two reasons: first, we have population appears to segregate roughly into two groups: shown a small but significant difference in survival time 40% of the F2 mice are as resistant as the B6 parent and between the C3Ou and C3 strains, and second, the C3 survive the infection, whereas the rest of the F2 mice strain is known to carry a nonfunctional allele of the Tlr4 succumb following infection, but display a significantly locus.21,22 Tlr4-mediated responses have previously been spread and delayed time of death compared with their implicated in the onset of tissue pathology and morbidity C3 parents. C3OuB6F1 mice displayed an intermediate in the context of C. rodentium infection.20 Separate but fairly resistant phenotype, with an 81% survival rate analyses would enhance our ability to distinguish by day 30 post infection (Figure 2b). Also, 199 progeny common as well as distinct genetic factors controlling from a (C3Ou Â B6)F2 cross (hereafter named C3OuB6F2) survival to C. rodentium infection in the C3Ou and C3 showed a continuous distribution in survival very mouse strains. comparable with the one displayed by the C3B6F2

70 (C3 x B6) F2 60

30 Frequency 20

0 8 10 12 14 16 18 20 22 24 26 28 30 Survival time (days)

120 (C3Ou x B6) F2 110 100 90 80 70 60 50 Frequency 40 30 20 10 0 8 10 12 14 16 18 20 22 24 26 28 30 Survival time (days) Figure 2 Cumulative survival in the F1, F2 and the parental strains following gastric gavage with 2.5 Â 108 CFUs of C. rodentium. (a) (C3H/HeJxC57BL/6J)F2 cross. In all, 168 F2 animals were infected and survival was recorded. The F1 mice behaved as intermediates between the resistant C57BL/6J and the susceptible C3H/HeJ parents, with roughly 50% survival by day 30 post infection. The F2 population appears to segregate more or less into two groups. Approximately 50% of the F2 mice are as resistant as the B6 parent and survive the infection, whereas the rest of the F2 mice succumb following infection but display a significantly delayed and spread time of death compared with their C3 parents. (b) (C3H/HeOuJ Â C57BL/6J)F2 cross. In all, 199 F2 animals were infected and survival was recorded. The F1 mice displayed an intermediate but fairly resistant phenotype, with an 80% survival rate by day 30 post infection. The F2 mice showed a continuous distribution in survival very comparable with the one displayed by the C3 F2 cross except slightly shifted toward a more resistant phenotype. (c) Examination of the frequency distribution of the survival in the (C3 Â B6)F2 mice reveals that the phenotype is not normally distributed and presents a major spike in the distribution at day 30 (censored data, shown with crosshatched pattern). Also, when considering only nonsurviving mice (noncensored data), the phenotype is not normally distributed as there is a sharp increase in mortality early in infection (days 10–14) and then a gradual decrease in mortality up to days 24–27, leading to a distribution with a long ‘tail’ toward the right-hand side of the graph. (d) Examination of the frequency distribution of the survival in the (C3Ou Â B6)F2 mice reveals an even larger spike in the distribution at day 30 (censored data). For the nonsurviving mice, the phenotype is not normally distributed, as seen for the C3-derived mice.

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 283 population. However, the C3OuB6F2 cumulative survival Localization of loci controlling the survival phenotype curve was slightly shifted toward a more resistant The 168 C3B6F2 and the 199 C3OuB6F2 mice were each phenotype (58% survival at day 30) than that of C3B6F2 genotyped for polymorphic markers covering their entire mice. An examination of the frequency distribution for genome. These markers were all chosen for their ability the survival time of C3B6F2 mice (Figure 2c) reveals that to discriminate between B6 and C3H chromosomal the phenotype is not normally distributed. There is a segments being inherited by the F2 hybrids at each major spike in the distribution profile at day 30 locus. Supplementary Table 1 shows the names and representing surviving mice (censored data, shown with positions of the markers used. The genetic maps derived crosshatched pattern). The survival time distribution for from the crosses in this study are shown in Supplemen- C3OuB6F2 mice reveals an even larger spike of surviving tary Figure 2. Given the nature and distributions of the mice at day 30 (Figure 2d). When considering only phenotypes under study, a parametric survival regres- nonsurviving mice (noncensored data), the phenotype is sion on the markers, under the Cox proportional hazards not normally distributed either: there is a sharp increase model, was chosen for the genome scan and performed in mortality early in infection (days 10–14) and then a with the statistical program R (http://www.R-project. gradual decrease in mortality up to days 24–27, leading to org). The P ¼ 0.05 genome-wide significance threshold a distribution with a long ‘tail’ toward the right-hand side for the LOD (logarithm (base 10) of odds) scores, of the graph. This is seen in both crosses, although it is obtained by interpolation of the P-values computed more evident in the C3 cross (Figures 2c and d). using 100 000 bootstrapped samples, was 3.57 for the Importantly, no sex-dependent differences were observed (C3 Â B6)F2 cross and 3.28 for the (C3Ou Â B6)F2 cross. among the parental inbred strains or among cross One single locus was found to be associated with progeny (data not shown). All analyses shown in this survival in the (C3 Â B6)F2 cross (Figure 3a): a highly study are therefore the result of pooling male and female significant LOD score of 13.42 (Po10À5) was detected at data. position 13.58 cM on proximal chromosome 15 (chr15).

6 LOD score

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Genetic Markers

6 LOD score

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Genetic Markers Figure 3 Parametric survival regression at the genetic markers under the Cox proportional hazards model for the (C3 Â B6)F2 mice (a) and the (C3Ou Â B6)F2 mice (b). The LOD scores are shown for each position across the genomes. The horizontal dashed lines represent the genome-wide significance threshold using 100 000 bootstrapped samples (P ¼ 0.05). One significant association is found for the (C3 Â B6)F2 cross (Cri1 on mChr15) and two significant associations are detected for the (C3Ou Â B6)F2 cross (Cri1 on mChr15 and Cri2 on mChr5).

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 284 Table 1 LOD scores and P-values for survival analysis under the Cox regression model for the (C3 Â B6)F2 and (C3Ou Â B6)F2 mice

JAX SNP ID rs no. C3 C3Ou (chromosome-build30position-source) LOD score P-value LOD score P-value

05-116192311-M rs3661429 0.024 1.000 3.489 0.031

WI_WGS_15_4921676 rs3698904 7.126 o10À5 4.205 0.006 15-005994001-M rs3715343 7.023 o10À5 3.610 0.024 15-020953071-M rs3714893 10.118 o10À5 4.478 0.003 15-028279166-G rs13482497 12.882 o10À5 7.331 o10À5 15-033125499-M rs3720676 13.422 o10À5 9.341 o10À5 15-039906196-M rs3683812 13.368 o10À5 12.556 o10À5 15-044071932-M rs3694011 12.248 o10À5 12.164 o10À5 15-049283927-M rs3719583 8.656 o10À5 9.202 o10À5 15-057561875-M rs3702158 6.159 o10À5 7.885 o10À5 15-063090495-M rs3656705 6.230 o10À5 6.803 o10À5 15-070911071-M rs3703015 4.082 0.007 6.303 o10À5 15-075635928-N rs4230816 2.886 0.109 5.170 0.001

Abbreviations: LOD, logarithm (base 10) of odds; SNP, single-nucleotide polymorphism. All results with a P-value o0.05 (using 100 000 bootstrapped samples) are listed. The genetic marker most strongly associated with survival in each of the two crosses is shown in bold. The most proximal marker used within chromosome 15 was derived from the WICGR2 (Whitehead/MIT Center for Genome Research) SNP panel that uses a different nomenclature from the KBiosciences panel used for all other markers.

This locus will be referred to as Cri1 (Citrobacter cumulative survival curve that was not significantly rodentium infection 1). The statistics in Table 1 define a different from that of homozygous C3 mice, with a final genetic interval of significant association with survival survival rate of 28%. The same analysis is shown for the with LOD score peak at marker 15-033125499-M (C3Ou Â B6)F2 cross in Figure 4b. In that cross, 96% of the (rs3720676), which is physically located at chr15: mice that inherited both Cri1 alleles from the B6 parent 32 963 563 bp according to build 37 of the mouse genome survived, whereas only 30% of the F2 progeny with a assembly. The 1.5-LOD support interval for this peak homozygous C3Ou allele at Cri1 survived. In contrast spans 28.1 Mb, from marker 15-020953071-M (rs3714893) with the C3B6F2 mice, however, heterozygotes (53% to marker 15-049283927-M (rs3719583). survival) were significantly more resistant than homo- The (C3Ou Â B6)F2 cross (Figure 3b) yielded two loci zygous C3Ou littermates (P ¼ 0.02). Both of these graphs that are significantly associated with the survival show clearly that C3H-derived Cri1 alleles confer suscept- phenotype. A highly significant LOD score of 12.56 ibility to C. rodentium infection in terms of survival. (Po10À5) was detected at position 12.63 cM on chr15. However, whereas the C3H allele is clearly dominant in That peak overlaps almost perfectly with the Cri1 locus the context of the C3 cross, it appears co-dominant or found in the C3 cross as can be seen in Table 1. The peak additive in the C3Ou cross. Once again, it appears that of the genetic interval with significant association with C3Ou mice produce progeny that are consistently more survival occurs at marker 15-039906196-M (rs3683812), resistant to infection than those from the C3 strain. which is physically located at chr15: 39 561 915 bp C3OuB6F2 mice were also stratified according to their according to build 37. This peak has a 1.5-LOD support genotype at the Cri2 locus on chromosome 5 (Figure 4c). interval spanning 15.9 Mb, from marker 15-033125499-M The animals that inherited both copies of marker (rs3720676) to marker 15-049283927-M (rs3719583). 5-116192311-M from their B6 parent displayed a 35% In the (C3Ou Â B6) F2 cross only, a second significant cumulative survival rate at day 30 post infection that is peak was detected in the middle of chromosome 5, at significantly lower (P ¼ 0.008) than littermates that position 50.69 cM with the LOD score of 3.49 (P ¼ 0.03; inherited one or two copies from their C3Ou parent (68 Cri2). That peak corresponds to marker 5-116192311-M and 61% survival, respectively). Thus, the C3Ou allele of (rs3661429) and there is no flanking marker with the the Cri2 locus confers a modest level of resistance to LOD score above the 3.3 genome-wide significance C. rodentium infection and is dominant over the B6 allele. threshold defined above. No such peak can be observed in the (C3 Â B6)F2 cross. Tlr4 does not control the survival phenotype under study C3B6F2 mice used in this study effectively segregate the Phenotypic effect of the Cri1 and Cri2 loci Tlr4 deficiency of C3H/HeJ inbred mice. This enabled us A cumulative survival curve is shown in Figure 4a, with to clearly assess whether Tlr4 has any effect on mouse the C3B6F2 mice grouped according to their genotype at survival to C. rodentium infection. The genome scan the peak marker for the Cri1 locus. Most mice that shown in Figure 3a clearly fails to detect any significant inherited two B6-derived copies of the Cri1 locus linkage between mouse chromosome 4, where the Tlr4 survived the infection (94% survival at day 30 post gene is located, and the survival phenotype under study. infection), whereas mice that inherited two C3-derived To confirm that Tlr4 does not impact survival in our copies of the Cri1 locus displayed a 20% survival rate. C3B6F2 cross, we analyzed the survival of all 168 mice in Mice with a heterozygous genotype at Cri1 displayed a this cross according to their genotype at the Tlr4 gene

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 285 100 Tlr4 (chr4) 100 (C3 x B6) F2 C Cri1 (chr15) 80 H 80 (C3 x B6) F2 B C 60 60 H B 40 40 Percent survival Percent survival 20 20

0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Day post infection Day post infection Figure 5 Cumulative survival following C. rodentium infection in Cri1 (chr15) (C3 Â B6)F2 mice according to their genotype at Tlr4. Mice were (C3Ou x B6) F2 100 grouped according to their genotype at genetic marker rs3023006 that represents the H712P amino-acid substitution in the Tlr4 C protein of C3H/HeJ inbred mice responsible for their lipopolysac- 80 H charide (LPS)-unresponsive phenotype. ‘B’ represents mice homozygous for the B6 allele. ‘C’ represents mice homozygous for the B 60 C3H allele. ‘H’ represents mice heterozygous at the typed marker.

40 (Supplementary Figure 3). No significant effect of Tlr4 Percent survival 20 genotype on survival was seen, even in these stratified data. As discussed above, the Cri1 genotype of these mice was the major determinant of survival. 0 0 5 10 15 20 25 30 Day post infection Transfer of the resistance allele of Cri1 via congenic mice In order to study the effect of Cri1 in the absence of other 100 Cri2 (chr5) genetic differences between the parental strains, we (C3Ou x B6) F2 generated a congenic mouse strain that we named C C3Ou.B6-Cri1. This congenic strain carries a proximal 80 H 60 Mb segment of B6 chr15 encompassing the Cri1 locus B introgressed in the genetic background of the C3Ou 60 strain after at least nine backcross generations. Genotypic analysis of the congenic segment determined that it contains marker WI-WGS-15-4921676 (rs3698904; 0.5 Mb) 40 at the proximal side. As this marker is close to the

Percent survival telomere, it is likely that the congenic segment extends to 20 the telomere on the proximal side. On the distal side, marker (rs3702158; 57 Mb) is included and marker 15- 0 063090495-M (rs3656705; 62 Mb) is excluded from the 0 5 10 15 20 25 30 congenic segment. Therefore, the congenic segment in Day post infection these mice comprises at least 57 Mb and o62 Mb of Figure 4 Cumulative survival following C. rodentium infection in proximal chr15 from B6 mice introgressed onto the C3Ou (C3 Â B6)F2 mice (a) and (C3Ou Â B6)F2 mice (b) and according to background. C. rodentium infection of the congenic strain their genotypes at Cri1 and Cri2 (c). Mice were grouped according to followed by survival analysis was conducted. We found their genotype at the typed genetic marker most closely associated that the congenic strain thus created survived the with the peak LOD score on: (a) chromosome 15 (15-033125499-M), infection (96% cumulative survival). Effective transfer (b) chromosome 15 (15-039906196-M) and (c) chromosome 5 (5- 116192311-M). ‘B’ represents mice homozygous for the B6 allele. of the survival phenotype of the B6 inbred strain onto ‘C’ represents mice homozygous for the C3H allele. ‘H’ represents the otherwise susceptible C3Ou strain indicates that mice heterozygous at the typed marker. the genetic factor mediating the Cri1 effect is contained in the congenic region under study (Figure 6). Moreover, heterozygote congenic littermates tested in parallel (Figure 5). All three resulting survival curves were enabled us to show a complete dominance of the completely superimposable, thus clearly dismissing any C3Ou susceptibility allele of Cri1 (0% cumulative implication of the Tlr4 gene in the control of survival to survival at day 30 post infection). These heterozygous C. rodentium infections. To determine if the strong mice were not significantly different from noncongenic effect of Cri1 was masking a minor effect of Tlr4 on littermates and served as controls for the efficiency of the survival, the same 168 C3B6F2 mice were analyzed for infection experiments. Thus, Cri1 controls survival survival according to their genotype at Tlr4, but were following C. rodentium infection, and susceptibility is also stratified according to their genotype at Cri1 the dominant trait.

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 286 Cri1 does not control C. rodentium proliferation in the colon The congenic mice we created provide the ideal tool to The experiments described above clearly show that the determine if the other phenotypic differences described Cri1 locus controls mortality following C. rodentium between the parental strains are correlated with their infection. There are, however, a number of different increased mortality or not. One such phenotype is the phenotypes that can be studied during the infection and C. rodentium colonization of the intestinal tract of the that distinguish the C3H from the B6 mouse inbred host (days 1–10 post infection). strains, some of which have been shown in Figure 1. In order to test if the C. rodentium colonization difference between the B6 and C3Ou mouse strains is controlled by the Cri1 locus, we measured both bacterial shedding in fecal pellets (Figure 7a) and colonic CFU 100 counts (Figure 7b) in C3Ou.B6-Cri1 mice at day 5 post infection. In this study we provide further evidence that 80 C3Ou.B6-Cri1 homozygote measuring bacterial shedding in fecal pellets (as was C3Ou.B6-Cri1 heterozygote shown in Figure 1 as well) gives a good insight into the 60 C3Ou non-congenic littermates colonization of the colon by C. rodentium. Both types of measurements show that homozygous congenic mice 40 displayed high CFU counts that were similar to the ones observed in control C3Ou mice and significantly higher Percent survival 20

10 0 0 5 10 15 20 25 30 Day post infection 8 Figure 6 Cumulative survival following C. rodentium infection in Cri1 congenic mice. C3Ou.B6-Cri1 congenic mice obtained after nine 6 backcross generations were intercrossed and either used directly for infection and survival analysis or homozygous congenic animals 4 identified and intercrossed for production of large numbers of homozygous congenic animals. A total of 28 homozygous C3Ou.B6- Cri1 animals were infected and survival was recorded. Only one of 2 these congenic animals died (at day 14 post infection) for a cumulative survival at day 30 post infection of 96%. Five Fecal bacterial counts (log cfu/g) heterozygote congenic and three noncongenic (essentially C3Ou 0 mice) littermates included in the experiments failed to survive B6 C3Ou C3Ou.B6-Cri1 C3Ou.B6-Cri1 beyond day 16 post infection. Thus, heterozygotes and noncon- parental parental homozygote heterozygote genics did not display statistically different survival curves and were both very significantly different from homozygous congenics 10 (Po0.0001 according to a log-rank test). 8

Figure 7 Further phenotyping of Cri1-congenic mouse hosts 6 during C. rodentium infection. (a) Bacterial shedding measured as C. rodentium CFU g–1 of homogenized fecal matter was measured at day 5 post infection. In all, eight homozygous congenics, four 4 heterozygous congenics and six mice of each parental strain were used for this experiment. The bar graph depicts mean CFU counts with s.d. for each mouse strain. CFU counts obtained from both 2

homozygous and heterozygous congenic mice were similar to those Colon bacterial counts (log cfu) obtained from C3H controls tested in parallel and statistically 0 different (higher) than those obtained from B6 controls (Po0.05 B6 C3Ou C3Ou.B6-Cri1 C3Ou.B6-Cri1 according to Dunn’s post tests). (b) Bacterial colonization measured parental parental homozygote heterozygote as C. rodentium CFUs per 1 cm sample of homogenized colon was measured at day 5 post infection. In all, 18 homozygous congenics, 8 0.4 heterozygous congenics and 12 mice of each parental strain were Uninfected used for this experiment. The bar graph depicts mean CFU counts with s.d. for each mouse strain. CFU counts obtained from both Infected homozygous and heterozygous congenic mice were similar to those 0.3 obtained from C3H controls tested in parallel and statistically different (higher) than those obtained from B6 controls (Po0.01 c according to Dunn’s post tests). ( ) Colon weights in mice at day 9 0.2 post infection with C. rodentium. Between three and six mice of each strain were used for this experiment. Uninfected mice or mice at

day 9 post infection were killed and their entire colons from cecum (g) weight Colon to rectum were emptied of fecal pellets and weighed. The bar graph 0.1 depicts mean weights with s.d. for each mouse strain/treatment. Colon weights obtained from infected congenic mice were similar to those obtained from C3Ou controls tested in parallel and 0.0 statistically different (higher) than those obtained from infected B6 B6 C3Ou C3Ou.B6-Cri1 controls or uninfected colons from any of the strains (Po0.01 parental parental homozygote according to Tukey’s post tests).

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 287 than the extremely low CFUs observed in B6 control mice their genotypes at both the Cri1 and Cri2 loci (Supple- infected in parallel (Po0.05 according to a Dunn’s post mentary Figure 5), it becomes evident that mice that are test). Thus, the bacterial proliferation phenotype of the heterozygous for Cri1 alleles can be either as susceptible Cri1 congenic reflects the C3Ou background of the mice as homozygotes for C3Ou-derived Cri1 alleles (35% rather than the B6-derived segment encompassing the survival at day 30 post infection), or of an intermediate Cri1 locus. Indeed, when we performed a time course to phenotype (59% survival) depending on whether they compare the bacterial shedding of C3Ou and C3Ou.B6- inherited both Cri2 copies from their B6 parent or at least Cri1 mice over the entire course of infection, no one C3Ou copy, respectively (significant effect of the Cri2 significant differences were seen between the two mouse locus with P ¼ 0.02). Interestingly, C3OuB6F2 mice with strains up until and including the days that C3Ou mice homozygous alleles at Cri1 were unaffected by their Cri2 succumbed to infection (days 1, 3, 5, 7 and 9; Supple- genotype. Our attempts to define locus interactions mentary Figure 4). By the day 14 time point, all C3Ou statistically were hampered by the relatively small mice had succumbed to the infection. Notably, all sample size of the F2 crosses used. Given the very high C3Ou.B6-Cri1 congenic mice survived the infection LOD scores obtained at the Cri1 locus and the modest despite displaying similar colonization kinetics and effect of any other loci, significantly larger sample sizes degree as C3Ou mice. Therefore, the increased early would be required to detect interactions reliably. Ulti- bacterial proliferation seen in C3H strains of mice is not mately though, we decided to focus our efforts on controlled by Cri1 and is unrelated to their mortality confirming and isolating the effect of the Cri1 locus after infection. through the generation of congenic animals. To determine the eventual outcome of infection in As opposed to the data acquired for the F2 cross, the C3Ou.B6-Cri1 mice, we followed the fecal shedding of congenic data that we have generated point to a C. rodentium in C3Ou.B6-Cri1 mice beyond the time point complete dominance of the C3Ou-derived Cri1 allele in at which C3Ou mice die and determined that C3Ou.B6- conferring susceptibility to C. rodentium infection. To- Cri1 mice progressively clear the infection (Supplemen- gether, these results suggest that there were factors being tary Figure 4). By day 29 of infection, there were no segregated in the F2 cross (other than Cri2) that have an detectable C. rodentium in the colons of 3/6 C3Ou.B6-Cri1 impact on the survival phenotype and that have been mice and only low amounts (under 2.9 Â 103 CFU g–1)in eliminated with the congenic approach. the other three mice (data not shown), providing strong It has been previously suggested that it is the adaptive evidence that these mice clear the infection completely. immune system, crypt cell apoptosis and bacterial translocation but not lipopolysaccharide responsiveness Cri1 does not control C. rodentium colonic hyperplasia that may contribute to the tissue pathology and mortality As discussed above, the degree of colonic hyperplasia seen during C. rodentium infection of highly susceptible 19 during infection is much greater in C3 and C3Ou mice mouse strains. Our work has ascertained that the Tlr4 than in B6 mice. In order to quantitate the extent of gene does not control survival to this Gram-negative colonic hyperplasia during infection in the different extracellular bacterium. mouse strains, and to determine whether this difference Importantly, the Cri1 congenic mice generated in this is controlled by Cri1, we determined the colon weights at study have not only confirmed the localization of the day 9 post infection in B6, C3Ou and C3Ou.B6-Cri1 Cri1 locus to proximal chr15 but they have also enabled (Figure 7c). As shown in the graph, the colon weights of us to dissociate the survival phenotype controlled by the C3Ou.B6-Cri1 mice were indistinguishable to those of Cri1 locus from any control of C. rodentium proliferation C3Ou mice and significantly different from B6 mice, or colonization of the host’s distal colon. It has indicating that Cri1 does not control colonic hyperplasia previously been suggested that these two phenotypes 19 following C. rodentium infection. were related. However, our Cri1 congenic data clearly show that mice can survive the infection even in the context of rapid bacterial replication. Thus, Cri1 controls survival of the host independently of the ability of the Discussion pathogen to colonize the gut in the early phase of the We have applied a systematic approach to the identifica- infection. We have extended our genetic screening of tion of genetic factors responsible for the differential B6C3OuF2 mice to identify the loci that are responsible survival of B6 and C3H inbred mouse strains infected for the differential proliferation of C. rodentium in the with the enteric pathogen C. rodentium. A genome scan colon of these two mouse strains and have thus far been using F2 crosses has yielded one major autosomal locus unable to pinpoint the genetic loci responsible for a within proximal chr15 that we have named Cri1. This significant part of the phenotypic difference (data not locus explains B25% of the phenotypic variance among shown). We are therefore tempted to suggest that it may the F2 mice studied. Our working hypothesis is that the be under complex genetic and perhaps environmental Cri1 locus within chr15 is the main genetic determinant control. of survival to C. rodentium infection in both the C3H/HeJ The mortality of C3H inbred mice is a rather unique and C3H/HeOuJ mouse strains. The C3Ou-specific Cri2 phenotype among mouse strains as there are many locus within chromosome 5 could potentially explain the strains other than B6 that generally survive C .rodentium small but significant difference in survival observed infections (129S1, A/J, BALB/c, CBA/J, DBA/2J, between the C3 and C3Ou C3H mouse strains. It is Ma/MyJ, SW;19,23,24 in-house data not shown). There is, tempting to speculate that Cri2 may also be responsible however, one more mouse strain that has been shown to for the apparent difference in the mode of inheritance of display reduced survival during C. rodentium infection: the Cri1 locus between C3B6F2 and C3OuB6F2 progeny. the FvB/NTac strain.23 FVB mice have been character- Indeed, when C3OuB6F2 mice are stratified according to ized with significant weight loss and mortality beginning

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 288 by 9 days post infection and B75% mortality by 21 days generating subcongenic mice that transfer smaller por- post infection. The mice develop epithelial cell hyper- tions of the Cri1 locus from the B6 parental strain onto proliferation and severe inflammation with erosions and the C3Ou genetic background. Once a well-defined and ulcers in the descending colon by 6 days post infection. smaller genomic candidate interval has been identified, Mortality can be fully prevented by fluid therapy, and we will undertake the screening of candidate genes for genome-wide gene expression profiling of FVB mice coding nonsynonymous mutations as well as for reg- compared with resistant SW mice has revealed an ulatory mutations. Functional validation of prioritized association between mortality and impaired intestinal candidate genes in vivo will then be used to link ion transport and development of fatal fluid loss and candidate genes to the Cri1 survival phenotype. dehydration.25 Two ion transport genes were pinpointed because of their parallel expression profiles in the FVB and C3H mouse strains, namely a reduced expression by Materials and methods day 9 post infection, compared with uninfected mice as well as with infected B6 and SW resistant mice. The two Animals used genes identified in those studies were Slc26a3 (mouse All animal procedures were performed under conditions chromosome 12) and CAIV (Car4) (mouse chromosome specified by the Canadian Council on Animal Care. 11). Their findings suggested a common pathway C57BL/6J, C3H/HeJ and C3H/HeOuJ were purchased (chloride ion absorption) associated with mouse mortal- from the Jackson Laboratory (Bar Harbor, ME, USA). ity in both susceptible mouse strains.26 However, our All animals were maintained in a specific pathogen-free systematic approach to finding genetic loci that control facility at McGill University, and all experiments were the survival difference between B6 and C3H mice does approved by the animal care committee. (C57BL/ not identify either of these two ion transporters as major 6J Â C3H/HeJ) F1 and F2 and (C57BL/6J Â C3H/HeOuJ) causative agents. This finding, together with the fact that F1 and F2 mice were generated at the McGill University the downregulation of expression of these ion transport Lyman Duff animal facility. genes is only detectable by day 9 post infection, suggests that these expression differences might be the effects of In vivo C. rodentium infections lethal infection in the susceptible strains rather than Mice were orally inoculated either with C. rodentium causative factors. Our current study identifies Cri1 as a strain DBS100 wild type or that same strain rendered causative genetic agent for the mortality of C3H mice, resistant to chloramphenicol (Cm) by the insertion of a but a proper genetic analysis of FVB mortality should be mini-Tn7 transposon encoding chloramphenicol resis- performed before a conclusion is reached about whether tance into the glmS gene. The resulting antibiotic Cri1 controls survival of FVB mice as well. Notable resistant strain was shown to be indistinguishable from phenotypic differences between the FVB and C3H strains the parental strain in terms of infectivity and kinetics as determined by Borenshtein et al.26 included a during mouse infection.31 For inoculations, bacteria were significantly delayed mortality profile, slower body grown overnight in 3 ml Luria-Bertani broth shaking weight loss and lower bacterial shedding during the at 37 1C. Mice were infected by oral gavage of 0.1 ml early stages of infection for FVB mice. The studies of Luria-Bertani broth containing 2.5 Â 108 CFUs of performed by these authors are crucial in that they show C. rodentium. The infectious dose was verified by plating that fluid therapy effectively rescues the survival of C3H of serial dilutions on selective MacConkey plates. For and FVB mice and therefore places dehydration caused survival analysis, the mice were monitored one or two

by diarrhea as a probable cause of death. However, the times daily and moribund animals were killed with CO2. mechanism by which mortality is prevented in resistant Mice were killed if they met any of the following clinical mouse strains may or may not be dependent on ion end points: 20% body weight loss, hunching and transport activities. Candidate genes under the 1.5-LOD shaking, inactivity, ruffled fur, anal prolapse, overtly support interval for the Cri1 locus do include ion bloody stool, bleeding from the anus and body condition transport genes: the ATP synthesis-coupled proton score o2. Enumeration of the colon and fecal bacterial transporter Atp6v1c1 (38.6 Mb) and the two voltage- load was done on different days post infection as follows: gated potassium channel subunits Kcns2 (34.8 Mb) and the last centimeter of the colon or feces (5 to 6 freshly Kcnv1 (44.9 Mb). However, the Cri1 candidate interval collected pellets) were weighed and then homogenized can also be considered to contain genes with a known in 1 ml of sterile phosphate-buffered saline using a function in the innate immune system such as the polytron homogenizer. The polytron was cleaned exten- complement component 9 (C9; 6.4 Mb) or the scavenger sively with 70% ethanol and water between samples. receptor Enpp2 (54.7 Mb). It also overlaps the FYN- Homogenates were serially diluted (0.1 in 0.9 ml) in binding protein (Fyb; 6.5 Mb) for which deficient mice sterile phosphate-buffered saline and 0.1 ml aliquots of display reduced T-cell proliferation27 as well as the each serial dilution were plated on MacConkey agar interleukin 7 receptor (Il7r; 9.4 Mb), known to be an containing chloramphenicol. C. rodentium was distin- important regulator of B- and T-cell proliferation/ guished by its characteristic colony morphology on this differentiation.28 Although the Il7r gene seems relatively medium as described previously.12,13,19,20,23,30 Plates con- far proximal of the peak linkage region of the Cri1 locus, taining between 30 and 300 colonies were counted. When it is an intriguing candidate as B-cell function has bacterial loads were low, leading to the undiluted sample been found to be essential for C. rodentium infection plate having o30 colonies, the number of colonies on survival.19,29,30 this plate was counted. Bacterial loads were then Ultimately, however, a continued systematic genetic expressed as CFU g–1 of feces or colon. The F2 mice were approach may be required to identify the gene under- infected by groups of 200 mice each time. The parental lying the Cri1 locus. For that purpose, we are currently strains were included as controls with every F1 and F2

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 289 infection. All mice were B4 weeks old at the time of Conflict of interest infection. The authors declare no conflict of interest. Genotyping DNA was extracted from biopsies of mouse tails with Acknowledgements overnight digestion in lysis buffer and proteinase K, followed by a chloroform extraction. DNA concentra- This work was supported by a team grant from the tions were measured with Quant-iT DNA Assay Kit Canadian Genetic Disease Network/Gene Cure Founda- (Molecular Probes, Invitrogen, Burlington, ON, Canada) tion (SG and DM) and a CIHR Operating Grant and adjusted to 10 ng ml–1. Single-nucleotide polymorph- MOP89817 (SG). E Diez was the recipient of a Canadian isms were chosen from the Jackson Laboratories Association of Gastroenterology/Canadian Institutes of and KBiosciences (Hoddesdon, Herts, UK) collaborative Health Research-partnered (CAG-CIHR) postdoctoral mouse single-nucleotide polymorphism panel.32 The fellowship. complete list of genetic markers used in the genome scan is included in Supplementary Table 1. Single nucleotide polymorphism genotyping was performed References by KBiosciences. In-house PCRs done for mouse genotyping were sent to Genome Quebec (Montreal, 1 Brennan PC, Fritz TE, Flynn RJ, Poole CM. Citrobacter QC, Canada) for sequencing with selected genetic Freundii associated with diarrhea in a laboratory mice. Lab interval markers on chr15. PCR was performed using Anim Care 1965; 15: 266–275. 2 Muto T, Nakagawa M, Isobe Y, Saito M, Nakano T. Infectious standard touchdown PCR methods with primers ob- megaenteron of mice. I. Manifestation and pathological tained from the Mouse Genome Informatics website observation. Jpn J Med Sci Biol 1969; 22: 363–374. (www.informatics.jax.org). 3 Nakagawa M, Sakazaki R, Muto T, Saito M, Hagiwara T. Infectious megaenteron of mice. II. Detection of coliform Generation of congenic strains organisms of an unusual biotype as the primary cause. Jpn J To generate homozygous congenic strains, C57BL/6J Med Sci Biol 1969; 22: 375–382. were crossed to C3H/HeOuJ mice to generate (C57BL/ 4 Ediger RD, Kovatch RM, Rabstein MM. Colitis in mice with a high incidence of rectal prolapse. Lab Anim Sci 1974; 6J Â C3H/HeOuJ)F1 mice, which then were backcrossed 24: 488–494. to the parental C3H/HeOuJ. The resulting N2 mice with 5 Barthold SW, Coleman GL, Bhatt PN, Osbaldiston GW, Jonas the desired B6-derived segment of chr15 were selected AM. The etiology of transmissible murine colonic hyperplasia. by genotyping and used for subsequent backcrossing to Lab Anim Sci 1976; 26(6 Part 1): 889–894. the C3H/HeOuJ parental strain. Males and females 6 Mundy R, MacDonald TT, Dougan G, Frankel G, Wiles S. (backcross generation N9) that inherited the intact Citrobacter rodentium of mice and man. Cell Microbiol 2005; 7: C57BL/6J-derived congenic segment (with no detectable 1697–1706. recombination) were intercrossed. Then, the offspring 7 Deborah Chen H, Frankel G. Enteropathogenic Escherichia were genotyped to identify homozygotes for the desired coli: unravelling pathogenesis. FEMS Microbiol Rev 2005; 29: 83–98. segment, which were then intercrossed to maintain the 8 Nataro JP, Kaper JB. Diarrheagenic Escherichia coli. Clin line. Mice were housed in a pathogen-free environment. Microbiol Rev 1998; 11: 142–201. 9 Centers for Disease Control and Prevention (2009). http://www. Gross and pathology analyses cdc.gov/nczved/divisions/dfbmd/diseases/enterohemorrhagic_ Two mice from each genotype were infected by oral ecoli/technical.html. gavage as done for the survival analyses. At day 7 post 10 Manning SD, Motiwala AS, Springman AC, Qi W, Lacher DW, infection, a gross/clinical examination of the animals Ouellette LM et al. Variation in virulence among clades of Escherichia coli O157:H7 associated with disease outbreaks. was performed. All mice were alert and moving around Proc Natl Acad Sci USA 2008; 105: 4868–4873. normally in the cages. Body weights were measured 11 Serna AT, Boedeker EC. Pathogenesis and treatment of Shiga and signs of dehydration and diarrhea were noted. The toxin-producing Escherichia coli infections. Curr Opin Gastro- entire intestinal tract was dissected and analyzed by a enterol 2008; 24: 38–47. pathologist. 12 Deng W, Li Y, Vallance BA, Finlay BB. Locus of enterocyte effacement from Citrobacter rodentium: sequence analysis Statistical analysis and evidence for horizontal transfer among attaching and effacing pathogens. Infect Immun 2001; 69: 6323–6335. One locus mapping for the survival phenotype for both 13 Deng W, Puente JL, Gruenheid S, Li Y, Vallance BA, Vazquez (C57BL/6J Â C3H/HeJ) and (C57BL/6J Â C3H/HeOuJ) A et al. Dissecting virulence: systematic and functional F2 mice was performed using a parametric survival analyses of a pathogenicity island. Proc Natl Acad Sci USA regression at the markers under a Cox proportional 2004; 101: 3597–3602. hazards model for the survival time distribution. The 14 Fagundes Neto U, Ferreira Vde C, Patricio FR, Mostaco VL, significance was assessed via 100 000 bootstrapped Trabulsi LR. Protracted diarrhea: the importance of the resamples controlling the genome-wide type 1 error. enteropathogenic E. coli (EPEC) strains and Salmonella in its The 1.5-LOD support intervals were obtained by inter- genesis. J Pediatr Gastroenterol Nutr 1989; 8: 207–211. polation on the physical map. Statistical differences 15 Fagundes-Neto U, Kallas MR, Patricio FR. Morphometric study of the small bowel mucosa in infants with diarrhea due between pairs of survival curves of mice grouped by to enteropathogenic Escherichia coli strains. Hepatogastro- genotype were calculated using log-rank tests. Statistical enterology 1997; 44: 1051–1056. differences within CFU data were calculated using one- 16 Karch H. The role of virulence factors in enterohemorrhagic way analysis of variance followed by pairwise Dunn’s Escherichia coli (EHEC)–associated hemolytic-uremic syn- post tests. drome. Semin Thromb Hemost 2001; 27: 207–213.

Genes and Immunity Cri1 and susceptibility to C. rodentium infection E Diez et al 290 17 Tacket CO, Sztein MB, Losonsky G, Abe A, Finlay BB, 26 Borenshtein D, Schlieper KA, Rickman BH, Chapman JM, McNamara BP et al. Role of EspB in experimental human Schweinfest CW, Fox JG et al. Decreased expression of enteropathogenic Escherichia coli infection. Infect Immun 2000; colonic Slc26a3 and carbonic anhydrase IV as a cause of 68: 3689–3695. fatal infectious diarrhea in mice. Infect Immun 2009; 77: 18 Barthold SW, Osbaldiston GW, Jonas AM. Dietary, bacterial, 3639–3650. and host genetic interactions in the pathogenesis of transmis- 27 Peterson EJ, Woods ML, Dmowski SA, Derimanov G, Jordan sible murine colonic hyperplasia. Lab Anim Sci 1977; 27: MS, Wu JN et al. Coupling of the TCR to integrin activation by 938–945. SLAP-130/Fyb. Science 2001; 293: 2263–2265. 19 Vallance BA, Deng W, Jacobson K, Finlay BB. Host suscept- 28 Peschon JJ, Morrissey PJ, Grabstein KH, Ramsdell FJ, ibility to the attaching and effacing bacterial pathogen Maraskovsky E, Gliniak BC et al. Early lymphocyte expansion Citrobacter rodentium. Infect Immun 2003; 71: 3443–3453. is severely impaired in interleukin 7 receptor-deficient mice. 20 Khan MA, Ma C, Knodler LA, Valdez Y, Rosenberger CM, J Exp Med 1994; 180: 1955–1960. Deng W et al. Toll-like receptor 4 contributes to colitis 29 Maaser C, Housley MP, Iimura M, Smith JR, Vallance BA, development but not to host defense during Citrobacter Finlay BB et al. Clearance of Citrobacter rodentium requires B rodentium infection in mice. Infect Immun 2006; 74: 2522–2536. cells but not secretory immunoglobulin A (IgA) or IgM 21 Poltorak A, He X, Smirnova I, Liu MY, Van Huffel C, Du X antibodies. Infect Immun 2004; 72: 3315–3324. et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr 30 Bry L, Brenner MB. Critical role of T cell-dependent serum mice: mutations in Tlr4 gene. Science 1998; 282: 2085–2088. antibody, but not the gut-associated lymphoid tissue, for 22 Qureshi ST, Lariviere L, Leveque G, Clermont S, Moore KJ, surviving acute mucosal infection with Citrobacter rodentium, Gros P et al. Endotoxin-tolerant mice have mutations in Toll- an attaching and effacing pathogen. J Immunol 2004; 172: like receptor 4 (Tlr4). J Exp Med 1999; 189: 615–625. 433–441. 23 Borenshtein D, Nambiar PR, Groff EB, Fox JG, Schauer DB. 31 Wickham ME, Brown NF, Boyle EC, Coombes BK, Finlay BB. Development of fatal colitis in FVB mice infected with Virulence is positively selected by transmission success Citrobacter rodentium. Infect Immun 2007; 75: 3271–3281. between mammalian hosts. Curr Biol 2007; 17: 783–788. 24 Chen C-C, Louie S, McCormick B, Walker WA, Shi HN. 32 Petkov PM, Ding Y, Cassell MA, Zhang W, Wagner G, Sargent Concurrent infection with an intestinal helminth parasite EE et al. An efficient SNP system for mouse genome scanning impairs host resistance to enteric Citrobacter rodentium and and elucidating strain relationships. Genome Res 2004; 14: enhances Citrobacter-induced colitis in mice. Infect Immun 1806–1811. 2005; 73: 5468–5481. 33 LeBlanc PM, Yeretssian G, Rutherford N, Doiron K, Nadiri A, 25 Borenshtein D, Fry R, Groff E, Nambiar P, Carey V, Fox J et al. Zhu L et al. Caspase-12 modulates NOD signaling and Diarrhea as a cause of mortality in a mouse model of regulates antimicrobial peptide production and mucosal infectious colitis. Genome Biol 2008; 9: R122. immunity. Cell Host Microbe 2008; 3: 146–157.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity