In the current study, we used this dietary iron system to investigate whether immune resistance created in this setting is necessary for asymptomatic carriage of C. rodentium . We found that adaptive immunity is not necessary for the iron-mediated antivirulence mechanism; however, the adaptive immune response is necessary for the continued suppression of virulence and asymptomatic persistence of the pathogen. Furthermore, asymptomatic carriage of genetically virulent but not genetically attenuated C. rodentium confers protection from subsequent challenges with the parental virulent strain of C. rodentium . Virulence behavior of C. rodentium , but not virulence factor antigens, is necessary to induce protective immunity, leading to systemic, but not intestinal, antibody production. Last, we demonstrate that an avirulent ler mutant of C. rodentium used in the field has secondary mutations in genes involved in LPS biosynthesis. Our work reveals previously unknown insight into how asymptomatic infections can arise mechanistically and that physiological defenses cooperate with immune resistance to confer protection against lethal infections.

The mouse-specific attaching and effacing (A&E) pathogen,, causes infectious colitis, which mimics diseases caused by human-associated A&E pathogens, enterohemorrhagic(EHEC) and enteropathogenic(EPEC) ( 21 25 ). Susceptibility toinfection is dependent on the genetic background of the mice ( 26 27 ). Whereas C57Bl/6 (B6) mice are highly resistant to, resulting in mild colonic inflammation and clearance of the pathogen by 3 weeks postinfection,causes severe and acute intestinal injury in C3H mice ( 28 29 ). The ability ofto cause disease requires virulence genes within the locus of enterocyte effacement (LEE) pathogenicity island (PAI) which encodes for a type 3 secretion system and effectors. The LEE PAI is regulated by the master virulence regulator, Ler, which is required for bacterial attachment and remodeling of the intestinal epithelium ( 30 31 ). We previously investigated host metabolic adaptations toinfection that regulate disease severity using the C3H model ( 19 ). In mice fed a normal diet, LEE virulence factor expression begins immediately after infection, with continued expression at the time of death around day nine postinfection. We demonstrated that administration of an iron-enriched diet led to activation of host antivirulence defense mechanisms that involved transient insulin resistance and increased availability of glucose in the gut, which limited LEE virulence expression by the pathogen, resulting in asymptomatic, persistent carriage of the pathogen. This phenotypic attenuation persisted even after withdrawal of iron diet. This was followed by within-host evolution of genotypically attenuatedcharacterized by nonfunctional mutations within the LEE PAI leading to an apparent commensalism between the host and

The host defense response, in part, dictates the heath trajectory a host will follow upon infection with a pathogen. Animals have evolved two distinct infection defense strategies that can be classified by their effects on pathogen fitness ( 9 ), i.e., antagonistic versus cooperative host defenses. Antagonistic defense mechanisms protect the host by having a negative impact on pathogen fitness. This includes mechanisms of immune resistance and nutritional immunity that operate to eliminate and kill pathogens ( 10 ). Antagonistic defenses also include behavioral avoidance mechanisms that reduce the risk of transmission ( 11 12 ). Cooperative or physiological defense strategies mediate host adaptation to the infection, yielding an apparent truce between the host and pathogen ( 13 14 ). This includes disease tolerance mechanisms that enable the host to withstand the presence of a pathogen by limiting physiological damage without killing the pathogen ( 15 18 ). In addition, antivirulence mechanisms are host physiological responses that can reduce the virulence behavior of the pathogen during colonization ( 19 20 ). Since cooperative defenses sustain host health with a neutral to positive impact on pathogen fitness, it has been proposed that this defense strategy can promote asymptomatic carriage of pathogens ( 9 19 ). In a mouse model of infection with a lethal enteric pathogen, promoting antivirulence defenses resulted in the persistent asymptomatic colonization of the pathogen ( 19 ). Studies to date have largely considered cooperative and antagonistic defenses as two independent defense strategies, although it is possible that these mechanisms may synergize to promote host survival and asymptomatic infections.

For many infections, there is substantial variation in host susceptibility to developing disease. For asymptomatic carriers, a pathogen can infect, replicate, and transmit without causing clinical signs or symptoms of disease in the primary host ( 1 2 ). Susceptible individuals that develop disease either recover after their sickness phase or ultimately continue to decline in health. Pathogen virulence, or the ability to cause sickness, depends on the damaging factors of the microbe as well as the host response to that pathogen. Much of the focus of understanding host susceptibility to infectious disease has focused on the host genetic makeup, metabolic or immune status, diet, and the microbiome that result in deteriorating health trajectories ( 3 8 ). We have relatively little understanding of how asymptomatic infections occur mechanistically or how they contribute to host defense and susceptibility to future infections. Because asymptomatic carriers contribute to infectious disease transmission and host-pathogen coevolution, it is necessary to take a more holistic approach for our studies of infectious diseases and not solely focus on mechanisms resulting in sickness but also mechanisms of asymptomatic infections.

Model of how iron-induced metabolic adaptations promote phenotypic attenuation of, development of adaptive immunity, long-term asymptomatic carriage, and protection from subsequent infection. LEE pathogenicity is required to prime the immune system and promote immune protection. 1. Under control diet conditions, wild-typeexpresses Ler and LEE virulence factors facilitating attachment, invasion, colitis, and death. 2. Under dietary iron conditions, there is an increase in glucose availability in the intestine ( 19 ), which suppresses expression of LEE-encoded virulence factors resulting in a largely phenotypically avirulent population ofthat remains lumenally bound. A small proportion of the pathogen population expresses LEE virulence factors, attaches to the epithelium, and induces systemicIgG, which is necessary for the selective exclusion of phenotypically virulentthat arises after iron withdrawal as well as protection from subsequent challenges with wild-type pathogen. 3. When infected with the Δor Δstrain, there is no phenotypically virulent pathogen to adhere to the intestinal epithelium and induce a systemic IgG response. When animals are challenged with a secondary wild-typeinfection, they are susceptible due to the lack of–specific IgG.

There are several models to explain how- and LEE-encoded virulence factors can mediate the generation of serum IgG and protection against secondarychallenges. The first model is that the host recognizes specific virulence factor antigens that are expressed on virulentbut not the attenuated strains that are required for the generation of serum IgG. A second model is that LEE-encoded virulence factors enable the pathogen to invade the host niche, which is necessary for pathogen recognition and the generation of serum IgG. Our data argue against LEE-specific antigens as the sole mediator of protection and so we consider the later model. Both the Δstrain and the Δstrain are defective in attachment/invasion ( Fig. 5A ) ( 30 42 ). However, while the Δstrain is defective in the expression of virulence factors within the LEE PAI (fig. S6, E and F), the Δstrain is not defective in expression of LEE virulence genes (fig. S6, E and F) ( 30 ). We used these strains to discriminate against the requirement for LEE virulence factors and pathogen behavior for regulation ofvirulence and protection against secondary challenges. Mice that received Δfor their primary challenge were susceptible to secondary infections with the parental virulent strain, with ~70% succumbing to the secondary challenge which is comparable to the range of susceptibility we observe in mice that received strains lacking LEE expression ( Figs. 2B and 6, C to E , and figs. S2I and S6, A and B). Furthermore, mice infected with the Δstrain fail to produce–specific serum IgG or lumenal IgA antibodies ( Fig. 6, A and B ). Together our data demonstrate that virulence activity or behavior is a prerequisite for long-term protection and asymptomatic carriage of phenotypically attenuated Fig. 7 ).

To determine the consequences of LPS and ler deficiency for an asymptomatic carriers’ ability to protect against subsequentchallenges, we infected C3H mice with wild-type,, ∆, or ∆strains or treated with PBS for mock infection and gave iron diet for 2 weeks. At 4 weeks postinfection with the primary challenges, we rechallenged the mice with a lethal dose of wild-type. Similar to Fig. 2B ), ∆also did not provide protective immunity ( Fig. 6, C to E ). Mice immunized with the LPS strains were completely protected from secondary challenge with the wild-type strain, demonstrating that the presence of LPS is not necessary for development of resistance defenses against subsequent virulent challenges ( Fig. 6, C to E , and fig. S6, A and B). These data also demonstrate that serum IgG responses correlate with potential for protective immunity, suggesting that serum IgG is important for defense againstwhile lumenal IgA is dispensable. In a complement bactericidal assay, serum from iron fed mice and infected with wild-typekills the pathogen ( Fig. 6F and fig. S6C). By contrast, unlikeand muMt mice,mice are completely protected from infection with wild-type(fig. S6D). Thus, ler- and LEE-encoded virulence factors are necessary for resistance against secondary challenges in asymptomatic carriers of, via the induction of serum IgG responses.

and) Whole bacteria ELISAs quantifying (A) lumenal IgA or (B) serum IgG against wild-typeAntibody samples were collected at 2 weeks postinfection from C3H mice infected with PBS (mock), wild-type, ∆, orand fed iron diet.= 8 to 10 mice per condition. Data represent one biological replicate. (to) Secondary challenge experiments were performed as depicted in Fig. 2A but with primary infections of PBS (mock), wild-type, ∆, orand secondary infections with Cm-resistant wild-type(C) Survival, (D) percent original weight, and (E) fecal shedding were monitored.= 6 to 10 mice per condition. Data represent two biological replicates combined. () Complement bactericidal assay using serum from mock-infected or wild-type–infectedmice fed iron. Serum was collected at 14 days postinfection, and killing was tested on wild-type= 3 mice per condition. One-way ANOVA with post-Tukey test, Kruskal-Wallis with post-Dunn’s test, two-way ANOVA, or unpairedtest for pairwise comparisons. Log rank analysis for survival. Error bars are ±SEM for (D) and geometric mean ± geometric SD for (E) and (F). *< 0.05, **< 0.01, ***< 0.001, and ****< 0.0001.values in (C) to (E) are shown condition versus wild type.

Mice fed dietary iron and infected with thestrain ofare not protected from subsequent challenges with the parental virulent strain of the pathogen ( Fig. 2 ). While our studies with the ΔLEE strain (fig. S2I) suggest that LEE virulence factors are required during the primary infection for the protection of asymptomatic carriers against subsequent challenges, because we found theto have background mutations in genes important for LPS synthesis, we wanted to rigorously test a role for LPS in addition to Ler for mediating this protective effect using our clean deletion strains. First, we tested the importance of LPS and Ler for the generation of–specific antibodies during infection.andmice were orally infected with wild-type, the LPS defective Δ, or the Δstrain and given iron chow. We then collected lumenal and serum antibodies from mice at 14 days postinfection and tested their ability to bind specifically to wild-type. Mice infected with Δwere unable to generate gut IgA antibodies that recognize wild-type Fig. 6A ); however, ∆-infected mice did generate–specific serum IgG antibodies ( Fig. 6B ). Mice infected with Δdid not generate lumenal IgA or serum IgG antibodies that bind to wild-type Fig. 6, A and B ).

We next tested the effects of LPS deficiency forvirulence in SCID-deficient mice with our dietary iron model. Similar to mice infected with the wild-typestrain, a 2-week course of dietary iron was sufficient to protectmice from lethality and weight loss when infected with the Δstrain ( Fig. 5E and fig. S5, C and D). By contrast,mice succumb to infection with Δdespite receiving the 2-week course of dietary iron ( Fig. 5E and fig. S5, C and D). Instead, we found that infection ofmice with the Δstrain is attenuated ( Fig. 5E and fig. S5, C and D). Together, our data demonstrate that the adaptive immune and humoral response is necessary for defense against Ler-dependent virulence; however, this does not require antibody binding to Ler-encoded antigen as has been previously suggested.

( A ) Representative images of fluorescent actin staining assay of wild-type, ∆ ler , ∆tir , ∆wfaP , and ∆rfaK mutants attaching to HeLa cells. Green signifies host cell actin, and red signifies C. rodentium. White arrows denote foci of actin associated with bacteria. ( B to D ) C3H mice were infected with 7.5 × 10 8 CFU of wild-type, ∆ rfaK , ∆wfaP, and complement strains. (B) Survival, (C) percent original weight, and (D) fecal shedding. n = 9 to 10 mice per condition. Data represent one biological replicate. ( E ) SCID +/+ or +/− and SCID −/− mice were infected with wild-type, ∆ ler , or ∆rfaK and given iron diet for 2 weeks after which mice were switched back to their normal chow diet. Survival was determined. Wild-type ( SCID +/+ or +/− ) = 5 mice, wild-type ( SCID −/− ) = 3 mice, Δ ler ( SCID +/+ or +/− ) = 2 mice, Δ ler ( SCID −/− ) = 3 mice, Δ rfaK ( SCID +/+ or +/− ) = 9 mice, Δ rfaK ( SCID −/− ) = 7 mice. Data represent two biological replicates combined. Log rank analysis for survival. two-way ANOVA for pairwise comparisons. Error bars indicate ±SEM for weight curves and geometric mean ± geometric SD for CFU analyses.

We next tested the hypothesis that the adaptive immune system is necessary to control LPS-dependent virulence rather Ler/LEE-dependent virulence. We first tested the importance of LPS forvirulence in vitro by determining whether the LPS strains were defective in pedestal formation, a requirement for in vivo virulence ( 42 ), on HeLa cells. Both ∆and ∆induced pedestals but not attenuated controls, ∆or ∆a mutant lacking the translocated intimin receptor () required for cell attachment ( Fig. 5A ) ( 42 ). In vivo, infection of mice fed a normal chow diet with the ∆strain resulted in a mild delay in host weight loss and death kinetics compared to mice infected with the wild-type strain of. Analysis of pathogen burdens revealed no significant differences in peak pathogen burdens, although there was a delay in rate at which peak levels were reached in the ∆-infected mice ( Fig. 5, B to D , and fig. S5, A and B). Complementation of the ∆strain with a wild-type copy of the gene restored virulence of the strain ( Fig. 5, B to D , and fig. S5A). Infection with the ∆strain exhibited a slightly greater attenuated phenotype, with only 60% of infected mice succumbing to the infection and also protection from clinical signs of disease ( Fig. 5, B to D , and fig. S5, A and B). Analysis of the fecal pathogen burdens revealed that the ∆strain reached comparable peak levels as the virulent strains; however, the rate at which peak levels were reached was slower, indicating that LPS is necessary forvirulence in vivo. Thus, with respect to the wild-type strain,defective in LPS is mildly attenuated yet is still capable of causing a lethal infection in mice under normal chow conditions.

( A ) LPS quantification of E. coli , wild-type, ∆ ler , ler::Km , ler::Km + wfaP , ler::Km + rfaK , and ler::Km + wfaP + rfaK. ( B and C ) Whole bacteria ELISA quantifying (B) lumenal IgA or (C) serum IgG binding against wild-type, ∆ ler , ler::Km , ler::Km + wfaP , ler::Km + rfaK , and ler::Km + wfaP + rfaK. n = 12 samples per condition for (B) and n = 16 samples per condition for (C). Data represent one biological replicate. ( D ) LPS quantification of E. coli , wild-type, ∆ wfaP , ∆ wfaP + wfaP , ∆rfaK , and ∆rfaK + rfaK. ( E and F ) Whole bacteria ELISA quantifying (E) lumenal IgA or (F) serum IgG binding against wild-type, ∆ wfaP , ∆wfaP + wfaP (E), ∆rfaK , and ∆rfaK + rfaK (F) . n = 5 to 19 samples per condition. Data represent one biological replicate. Lumenal and serum samples were collected from wild-type C. rodentium –infected C3H SCID mice fed iron diet for 2 weeks postinfection. Statistical significance was calculated using one-way ANOVA with post-Tukey test, Kruskal-Wallis with post-Dunn’s test, or two-way ANOVA. * P < 0.05, ** P < 0.01, *** P < 0.001, and **** P < 0.0001

To examine the role of theandtransposon mutations on LPS structure and antibody recognition in thestrain, we constructed single and doubleandcomplements. LPS extractions and LPS blots revealed thatbut not wild-type or our newly constructed Δmutant is defective in LPS assembly ( Fig. 4A ). Double complementation ofandwas required to restore LPS in thestrain ( Fig. 4A ). This demonstrates that through some unknown events, thestrain had acquired two inactivating mutations within the same LPS biosynthesis pathway. We next asked whether these mutations were responsible for loss of antibody binding of thestrain. Complementation of both theandmutations inrestored–specific IgA recognition ( Fig. 4B ).had partially enhanced, although not significant, IgG binding compared to wild-type, but binding was diminished when complemented withand Fig. 4C ). This suggests thatLPS may partially maskfrom IgG recognition. Last, we generated clean in-frame deletions ofandon the wild-typebackground. As predicted, both ∆and ∆single-gene knockout strains were defective in LPS synthesis ( Fig. 4D ) and complementation restored full LPS synthesis ( Fig. 4D ). When we examined antibody binding by whole bacterial enzyme-linked immunosorbent assays (ELISAs), we found that individual LPS mutations disrupted IgA binding to Fig. 4, E and F ). Together, these data demonstrate thatinfection with our dietary iron model results in the generation of pathogen-specific antibodies that require LPS and not-dependent antigens for binding.

Because we found intestinal IgA antibody binding was reduced with thestrain yet not restored with thecomplemented strain, it suggests that there are secondary mutations in thestrain resulting in antibody avoidance. We used this strain as a tool to investigate the IgA-dependent antibody recognition ofWe sequenced the genomes of wild-type,, and Δstrains, performed de novo hybrid genome assemblies, and identified secondary mutations in thestrain. Mutational analysis of thegenome identified multiple IS102 transposon duplications and insertions in ROD_21691 and ROD_41941, likely inactivating them (fig. S4A and table S1). Consistent with the genome analysis, polymerase chain reaction (PCR) amplifications of those genes inshowed PCR products +1000 bp greater than in either wild-type (fig. S4B). ROD_21691 and ROD_41941 are putative glycosytransferase and lipopolysaccharide (LPS) acetylglucosaminyltransferases, respectively, and herein renamedandfor their homology toLPS synthesis genes (fig. S4, C to E). WfaP is a predicted glucosyltransferase responsible for polymerization of O-antigen of LPS ( 40 ), whereas RfaK is essential to complete synthesis of the core subunit of LPS ( 41 ).

It has been previously reported that lumenal IgG–specific antibodies bind to LEE-encoded virulence factors and that this is necessary for the selective exclusion of virulentin the gut ( 33 ). We detected–specific lumenal IgA rather than lumenal IgG in our dietary iron system. Because we found that resistance against subsequent challenges ofrequires LEE genes in the primary infection and because previously challenged mice can selectively eliminate virulentin our dietary iron model, we posited that–specific lumenal IgA antibodies were recognizing LEE-encoded virulence factors in our system. From our antibody binding assays, we consistently saw that lumenal IgA isolated from wild-type–infected mice fed dietary iron had diminished recognition against, although binding was not completely ablated ( Fig. 3A and fig. S3, A and C). We also found that serum IgM antibodies isolated from wild-type–infected mice fed dietary iron had diminished recognition against, although binding was not completely ablated ( Fig. 3B and fig. S3, B and D). In addition, we found that serum IgG bound tostrain as well as to wild-type Fig. 3B and fig. S3, B and D) rather than a diminished response as previously reported forlacking 33 ). To determine whether the reduction inlumenal IgA was due to the loss ofand LEE virulence factor expression, we complemented themutation in thestrain. Although complementation partially rescued growth of the strain in mice (fig. S3, H and I), complementation did not restore gut IgA antibody binding to thestrain ( Fig. 3E ). We also examined lumenal IgA antibody binding to the ∆LEE strain. Consistent with our complementation studies, ∆LEE was still recognized by lumenal IgA at levels comparable to the wild-type strain (fig. S3J). Furthermore, we constructed a clean in-frame detection ofnamed ∆and tested it for antibody binding. The Δstrain did not evade lumenal IgA or serum IgG antibody binding ( Fig. 3E and fig. S3K) like thestrain. Together, these data suggest that discrimination between wild-type and attenuatedby antibodies is not dependent onor-regulated LEE-encoded virulence factors.

Whole bacteria ELISAs quantifying ( A ) lumenal or ( B ) serum IgA, IgG, or IgM antibodies binding to wild-type or ler::Km C. rodentium. n = 6 to 13 mice per condition. Data represent one biological replicate. ( C ) Survival and ( D ) percent original weight of wild-type, Rag −/− , and muMt − mice infected with 7.5 × 10 8 CFU wild-type C. rodentium. Mice were fed iron diet for 2 weeks and then placed on normal diet for the remainder of the experiment. n = 5 to 8 mice per condition. Data represent one biological replicate. ( E ) Whole bacterial ELISAs quantifying lumenal IgA binding against wild-type, ler::Km, ler::Km + ler , or ∆ ler. n = 5 to 19 mouse samples per condition. Data represent one biological replicate. Lumenal and serum samples were collected from wild-type C. rodentium –infected C3H SCID mice fed iron diet for 2 weeks postinfection. Mean ± SEM. Unpaired t test, Mann-Whitney test, or one-way ANOVA with post-Tukey test or two-way ANOVA was performed for pairwise comparisons. Log rank analysis for survival. ** P < 0.01, *** P < 0.001, and **** P < 0.0001.

Our experimental data thus far demonstrate that (i) adaptive immunity is necessary for asymptomatic colonization and long-term survival of the host in the iron-independent phase; (ii) mice develop a form of protective memory against subsequent challenges through selective exclusion of phenotypically virulentand (iii) development of this memory is dependent on LEE-encoded virulence in the iron-dependent phase. We reasoned that the mechanism of adaptive immune-mediated suppression of virulence may require antibody-mediated defenses and that this could be extended to provide protection against subsequent challenges with virulent. Afteroral infection, mice fed dietary iron for 2 weeks developed robust serum immunoglobulin G (IgG) and IgM, as well as intestinal IgA responses against wild-type Fig. 3, A and B , and fig. S3, A and B). Unlike previous reports ( 33 ), we were unable to detect–specific IgG in the gut of C3H or B6 mice ( Fig. 3A and fig. S3, A and C). As expected,or mock-infected mice did not develop–specific antibodies ( Fig. 3, A and B , and fig. S3, A to D). Last, and consistent with a role for antibody-mediated defense against 39 ), we found that muMT (B6 background) mice, which are deficient for mature B cells, were highly susceptible to an oral challenge with wild-typeinfection despite feeding of dietary iron ( Fig. 3, C and D , and fig. S3, E to G).

Although dietary iron limits the expression of the LEE virulence program to prevent detectable clinical and pathological effects of a primaryinfection on the host ( 19 ), because there is selective exclusion of virulentin response to the secondary challenge, we considered the hypothesis that virulence-competentis required in the primary infection to mediate resistance defenses against subsequent challenges with the same pathogen. To test this, we followed the experimental paradigm in Fig. 2A and challenged mice on iron diet with wild-typeorstrain for their primary challenge. In C3H/HeJ mice fed dietary iron, thestrain was not completely cleared, as indicated by persistent fecal shedding postprimary infection (fig. S2A). After 2 weeks, iron diet was removed and the mice were returned to control chow. At 4 weeks postinfection, mice were rechallenged with a lethal dose of wild-type Cm-resistant. A total of 100% of mice that received the primary challenge with virulentwere protected from the secondary challenge, but mice that receivedfor the primary challenge were susceptible to the wild-type secondary challenge ( Fig. 2, B and C , and fig. S2C). Similar to mock 1mice,mice susceptibility was associated with an inability to control Cm-resistantlevels ( Fig. 2D and fig. S2D). We also repeated these protection experiments with aisolate naturally missing the LEE PAI [herein called ∆LEE, isolate #9 from ( 20 )]. Likeprimary infection with ∆LEE strain also did not confer protection from secondary infection by wild-type(fig. S2I). Together, these data show that the presence of LEE-encoded virulence factors during the primary challenge in the asymptomatic host is required to mediate protective immunity against subsequent challenges of virulent

We next investigated how the host can accommodate persistent colonization withthat are phenotypically attenuated but can mount an effective resistance response against the secondary challenge. We hypothesized that the resistance response can selectively exclude phenotypically virulent pathogen. We tested this idea with a competition experiment as shown in Fig. 2E . We engineered different antibiotic tagged versions of wild-typeand obtained an attenuatedstrain,, which encodes a mutation in the master virulence regulator in 30 ).strains lackingdo not cause disease and are completely attenuated in mice ( 30 ). Compared to mice that are singly infected with the wild-type strain, mice singly infected with thestrain show reduced colonization when fed iron as indicated by fecal shedding levels. However, after 2 weeks postinfection when mice are returned to normal chow, the colonization levels are comparable to that found in mice singly infected with the wild-type strain (fig. S2, A and B). In vitro, growth of the wild-type and attenuated strain were comparable, demonstrating no inherent growth advantages exist (fig. S2E). We infected C3H/HeJ mice with a 7.5 × 10CFU dose of wild-typeand provided iron chow. After 2 weeks, iron chow was removed, and mice were returned to their normal chow diet. At 4 weeks postprimary challenge, we rechallenged mice with a 1:1 dose of wild-type and the attenuatedand continued to feed on normal chow. Although the attenuatedstrain displayed an initial competitive disadvantage, it quickly outcompeted the wild-type strain ofafter 1 week postinfection ( Fig. 2F and fig. S2F). The competitive advantage of thestrain was not observed in previously unchallenged mice (fig. S2, G and H). These data demonstrate that mice asymptomatically colonized with wild-typecan selectively exclude subsequent challenges with virulent but not LEE-attenuated

( A ) Schematic of secondary challenge experiments. C3H/HeJ mice were infected with of PBS (mock) or 7.5 × 10 8 CFU of wild-type or ler::Km and fed iron diet for 2 weeks. At 4 weeks postinfection, mice were rechallenged with 7.5 × 10 8 CFU of Cm-resistant wild-type C. rodentium. ( B ) Survival curve, ( C ) percent original weight postsecondary infection, and ( D ) fecal shedding of Cm-resistant C. rodentium . n = 7 to 19 mice per condition. Data represent three biological replicates combined. ( E ) Schematic of competition experiments. C3H/HeJ mice were infected with 7.5 × 10 8 CFU wild-type C. rodentium. Mice were fed iron diet between zero and 2 weeks postinfection. At 4 weeks postinfection mice, were challenged with different antibiotically tagged wild-type and ler::Km strains at a 1:1 ratio. ( F ) Competitive index scores over time calculated as wild-type or ler::Km over wild-type. n = 5 mice per condition. Data represent one biological replicate. Error bars are ± SEM for (C) and (F). Geometric mean ± geometric SD for (D). * P < 0.05, ** P < 0.01, and *** P < 0.001. Log rank analysis for survival. Unpaired t test or two-way ANOVA for pairwise comparisons.

Next, we determined the consequences of asymptomatic carriage of a pathogen for host defense against subsequent challenges with the same pathogen. The experimental paradigm is depicted in Fig. 2A . First, we challenged C3H/HeJ mice with an oral dose of phosphate-buffered saline (PBS) (mock 1) or an oral dose of 7.5 × 10colony-forming units (CFU) of wild-type(wild-type 1) which is normally a lethal infection in mice fed a normal diet ( Fig. 1B ) ( 19 ). Mice were provided iron diet for 2 weeks to induce asymptomatic carriage of the pathogen, with fecal pathogen burdens peaking at 10CFU/mg feces ( Fig. 2, to D , and fig. S2, A to C) ( 19 ). At 2 weeks postinfection with the primary challenge, the iron diet was removed and mice were returned to a normal chow diet. During this phase, pathogen burdens persisted at over 10CFU/mg feces (fig. S2, A and B). At 4 weeks postinfection with the primary challenge, we rechallenged the same mice with 7.5 × 10CFU of chloramphenicol (Cm)–resistant wild-type(2infection) to enable tracking of the secondary infection and kept mice on a normal chow diet. As expected, primary mock-infected mice given dietary iron were highly susceptible to the wild-type secondary challenge, with all mice exhibiting severe weight loss and an inability to resistgrowth, eventually succumbing to the infection ( Fig. 2, B to D , and fig. S2, C and D). By contrast, 100% of mice that were originally challenged with wild-typefor the primary round of infection (wild-type 1) and given dietary iron were protected from weight loss and survived the secondary challenge with Fig. 2, B and C , and fig. S2C). One possible model to explain this protection is that the host developed immune resistance against the secondary challenge to prevent colonization. Alternatively, following the primary infection, the host may sustain its antivirulence response which suppresses virulence of the secondaryinfection. Such a scenario might occur through sustained metabolic/physiological changes or sustained changes in the microbiome ( 32 ). To distinguish between these two models, we quantified the levels of Cm-resistantthat was being shed from primary mock- and wild-type–infected mice. In mock-infected mice, the levels of Cm-resistantreached 10CFU/mg feces by day 10 postinfection just before death. By contrast, there was a six log reduction in Cm-resistantbeing shed in the feces after the wild-type 1with a low level persisting by 3 weeks postinfection ( Fig. 2D and fig. S2D). Thus, asymptomatic carriage withseems to protect from subsequent lethal challenges via an antagonistic strategy, for example, by heightening the host resistance defenses.

) Schematic showing the dietary iron–mediated asymptomatic carriage ofand subsequent evolution of attenuation ( 19 ). In stage 1, dietary iron causes transient insulin resistance that is necessary to suppress expression of LEE PAI genes resulting in phenotypic attenuation. In stage 2, the host continues to remain asymptomatically colonized withthat is dependent on an iron-independent mechanism of phenotypic attenuation. In stage 3, an apparent commensalism between the host andhas formed due to within host evolution ofthat results in accumulation of mutations in LEE PAI genes. () survival, () percent original weight, and () fecal shedding of immunodeficient C3Hmice () or wild-type/heterozygous controls () infected with wild-typeAt the time of infection, mice were fed ad libitum control or iron diet for 2 weeks before returning to normal vivarium chow. Control diet= 11), control diet= 3), iron diet= 31), and iron diet= 13). Data represent two biological replicates combined. (C) Error bars ± SEM, (D) geometric mean ± geometric SD. Log rank analysis for survival and two-way analysis of variance (ANOVA) for pairwise comparisons. ns, not significant.

A schematic of our dietary iron model ofasymptomatic infection ( 19 ) is shown in Fig. 1A . We asked what host defense mechanisms are necessary for the continued suppression ofvirulence and asymptomatic carriage after removal of the iron diet. Specifically, we hypothesized that the adaptive immune system is necessary for asymptomatic carriage of. Thus, we generated immunodeficient SCID mice on the C3H/Snell background that are deficient for functional T and B cells () and challenged them with wild-typein the presence of iron rich diet or control chow.mice given dietary iron for 2 weeks had an initial survival advantage compared to bothandinfected mice fed a control chow diet ( Fig. 1B ). However 100% of the iron fedmice eventually died ~3 weeks postinfection (1 week post iron diet withdrawal), while iron fedwith functional adaptive immunity survived and were protected from infection-induced weight loss ( Fig. 1, B and C , and fig. S1A). Thus, the adaptive immune response is not necessary for iron-mediated protection from infection but is required for long-term survival during the iron independent phase. Examination of fecallevels over the course of the infection showed that there were comparable levels of pathogen inandinfected mice fed dietary iron at the onset ofdeath and with no significant differences in pathogen burdens at any other time point assessed in either the iron-dependent or iron-independent phase. This indicates thatmice have an impaired ability to adapt to the infected state ( Fig. 1D and fig. S1B). Thus, the adaptive immune response is necessary for the continued suppression of virulence and asymptomatic carriage of an enteric pathogen in an iron-independent manner.

DISCUSSION

C. rodentium, by promoting an antivirulence defense mechanism that prevents engagement of the pathogens virulence program without affecting its ability to colonize or replicate in the host (C. rodentium immunity (37, 38, C. rodentium to require a functional adaptive immune system. Our data support a model whereby the host transitions from antivirulence to immune resistance defenses to sustain asymptomatic persistence by selectively excluding phenotypically virulent pathogen, thereby, allowing phenotypically avirulent pathogen to persist ( Asymptomatic infections contribute to infectious disease transmission and host-pathogen coevolution. However, while we have made great advances in our mechanistic understanding of how pathogens cause disease, we have relatively no understanding of the mechanisms that facilitate the ability of pathogens to infect and replicate inside the host niche without causing disease. We previously demonstrated that dietary interventions can yield asymptomatic infections with the enteric pathogen, by promoting an antivirulence defense mechanism that prevents engagement of the pathogens virulence program without affecting its ability to colonize or replicate in the host ( 19 ). This phenotypic attenuation lasted for months after the removal of the dietary intervention, prompting us to ask whether host adaptive immunity contributes to the regulation of virulence and asymptomatic persistence in our model system. Consistent with the current knowledge aboutimmunity ( 34 43 ), we found long-term protection againstto require a functional adaptive immune system. Our data support a model whereby the host transitions from antivirulence to immune resistance defenses to sustain asymptomatic persistence by selectively excluding phenotypically virulent pathogen, thereby, allowing phenotypically avirulent pathogen to persist ( 44 ). We suggest that this is dependent on pathogen behavior and systemic IgG responses. In some contexts, it is logical to conceptualize cooperative defenses as a strategy to buy the host time for their adaptive immune response to kick in and clear the infection. Collectively, our studies demonstrate that indeed there is a cooperation between physiological defenses and adaptive immune resistance in defense against a lethal bacterial infection and that this can unexpectedly yield sustainment of an asymptomatic infection.

C. rodentium. Using a C57Bl/6 model, Kamada et al. (C. rodentium–infected mice that were phenotypically virulent and avirulent and that the adaptive immune response was necessary for the selective exclusion of virulent C. rodentium in the intestine. The authors proposed that gut IgG, and not IgA, binding of LEE-encoded virulence factor antigens that were expressed on virulent but not avirulent C. rodentium facilitated neutrophil-mediated elimination. In our studies, we were unable to detect C. rodentium–specific IgG in the gut and could only detect C. rodentium–specific IgA in the intestines of infected mice. We report that C. rodentium–specific gut IgA binding is not dependent on the presence of Ler/LEE virulence factors, and instead, binding is largely dependent on LPS. In any case, we demonstrate that gut IgA binding of LPS is not necessary for the selective exclusion of virulent C. rodentium. Instead, our data show that systemic C. rodentium–specific IgG mediates the selective exclusion of virulent C. rodentium. This is in agreement with previous studies from multiple groups demonstrating the importance of systemic IgG in host resistance against C. rodentium (33, tir mutant, we further demonstrate that virulent behavior (attachment/pedestal formation) mediated by LEE-encoded virulence factors rather than recognition of virulence factor antigens is what is necessary for selective exclusion of phenotypically virulent C. rodentium in the host niche. Our data support a model whereby systemic IgG is necessary to protect from phenotypically virulent C. rodentium that invades systemically after adhering to the intestinal epithelial niche (et al. ( Our data from the current study support a model that is mechanistically distinct from previous studies examining the interplay between the adaptive immune response, virulent and avirulent. Using a C57Bl/6 model, Kamada 33 ) also reported that there were subpopulations in–infected mice that were phenotypically virulent and avirulent and that the adaptive immune response was necessary for the selective exclusion of virulentin the intestine. The authors proposed that gut IgG, and not IgA, binding of LEE-encoded virulence factor antigens that were expressed on virulent but not avirulentfacilitated neutrophil-mediated elimination. In our studies, we were unable to detect–specific IgG in the gut and could only detect–specific IgA in the intestines of infected mice. We report that–specific gut IgA binding is not dependent on the presence of Ler/LEE virulence factors, and instead, binding is largely dependent on LPS. In any case, we demonstrate that gut IgA binding of LPS is not necessary for the selective exclusion of virulent. Instead, our data show that systemic–specific IgG mediates the selective exclusion of virulent. This is in agreement with previous studies from multiple groups demonstrating the importance of systemic IgG in host resistance against 34 ). Using the Δmutant, we further demonstrate that virulent behavior (attachment/pedestal formation) mediated by LEE-encoded virulence factors rather than recognition of virulence factor antigens is what is necessary for selective exclusion of phenotypically virulentin the host niche. Our data support a model whereby systemic IgG is necessary to protect from phenotypically virulentthat invades systemically after adhering to the intestinal epithelial niche ( Fig. 7 ). Some possible explanations for the discrepancies between our study and that by Kamada 33 ) are differences in the microbiome and strains of mice used that could complicate interpretations. Future work to understand the complexities of these relationships that can yield different results are needed.

ler::Km insertional mutants (sometimes confusingly annotated at Δler) that are used in the C. rodentium pathogenesis and immunity field (ler mutation), this strain acquired mutations in LPS genes. LPS mutants were partially attenuated in vivo, although LPS mutant–challenged mice were protected from secondary challenge against wild-type C. rodentium. It is likely that mutations in these LPS genes may also alter pathogen sensitivities against extracellular stress or host defenses (wfaP or rfaK resulted in reduced antibody recognition. In other Gram-negative bacteria, spontaneous mutations in rfaK can arise in the population, because of antibiotic stress or bacteriophages (ler::km strain the mutations occurred, we encourage investigators in the C. rodentium field to check their strains and, if appropriate, consider reassessing some of their phenotypes and conclusions. There are multipleinsertional mutants (sometimes confusingly annotated at Δ) that are used in thepathogenesis and immunity field ( 30 45 ). The strain used in this study is from ( 45 ) (see Materials and Methods). Our data show that at some unknown event in its lineage (before or after the generation of themutation), this strain acquired mutations in LPS genes. LPS mutants were partially attenuated in vivo, although LPS mutant–challenged mice were protected from secondary challenge against wild-type. It is likely that mutations in these LPS genes may also alter pathogen sensitivities against extracellular stress or host defenses ( 46 47 ). Single mutations inorresulted in reduced antibody recognition. In other Gram-negative bacteria, spontaneous mutations incan arise in the population, because of antibiotic stress or bacteriophages ( 48 ). Our work is a reminder of the importance to rigorously validate strains used in microbiological studies when making claims about gene functions including sequencing strains, complementing mutations and making clean deletions. As we do not know when in the lineage of thisstrain the mutations occurred, we encourage investigators in thefield to check their strains and, if appropriate, consider reassessing some of their phenotypes and conclusions.

et al. (C. rodentium eventually gets excluded from the intestine because it is outcompeted by the microbiota. However, in the current study, we demonstrate that phenotypically and genetically avirulent C. rodentium persists in the host as indicated by persistent fecal shedding. Differences in mouse stains, diets, and facilities may lead to differences in microbiota competitive effects with avirulent C. rodentium that can dictate whether the C. rodentium will persist or be outcompeted (50–C. rodentium (et al. (C. rodentium mouse infections, presumably through differences in the microbiota. Future work is needed to better understand how different microbiotas and mouse strains influence aspects of C. rodentium infection under normal and different dietary conditions including enriched iron. Kamada 49 ) demonstrated in C57Bl/6 mice that avirulenteventually gets excluded from the intestine because it is outcompeted by the microbiota. However, in the current study, we demonstrate that phenotypically and genetically avirulentpersists in the host as indicated by persistent fecal shedding. Differences in mouse stains, diets, and facilities may lead to differences in microbiota competitive effects with avirulentthat can dictate whether thewill persist or be outcompeted ( 32 53 ). In our previous work, we demonstrated in C3H/HeJ mice that dietary iron did not affect fecal shedding levels of 19 ). In the current study, we show that fecal shedding in iron fed mice is reduced compared to infected mice fed control diet in C3H/Snell mice. Furthermore, the work performed for the Sanchez 19 ) study was done in a different vivarium room than the current study, which further supports that in addition to mouse strain background, different environments can influence phenotypes ofmouse infections, presumably through differences in the microbiota. Future work is needed to better understand how different microbiotas and mouse strains influence aspects ofinfection under normal and different dietary conditions including enriched iron.