Salmonella SPI-1-mediated neutrophil recruitment during enteric colitis is associated with reduction and alteration in intestinal microbiota

Abstract

Gastrointestinal infections involve an interactive tripartite relationship between the invading pathogen, the host, and the host's resident intestinal microbiota. To characterize the host inflammatory response and microbiota alterations during enteric salmonellosis, C57BL/6 mice were pre-treated with a low dose of streptomycin (LD model) and then infected with S. typhimurium strains, including mutants in the two Type III secretion systems, SPI-1 and SPI-2 (invAmut and ssaRmut, respectively). Cecal colonization and inflammation in the LD model were evaluated to assess infection success and progression, and compared to the traditional high dose (HD) model. Perturbations to the microbial community in the LD model were assessed via evaluation of total microbial numbers, the proportion of intestinal γ-Proteobacteria and tRFLP analysis. In the LD model, consistently high colonization by the parental strain (WT) and invAmut S. typhimurium was associated with significant intestinal pathology. However, microbial community profiles were more similar both in numbers and composition between mice infected with the mutant strains, than with the WT strain. Consequently, significant infection-induced inflammation did not always produce similar microbiota perturbations. Large numbers of luminal neutrophils were observed in the ceca of WT-infected, but not in invAmut or ssaRmut infected mice. Neutrophils were thus implicated as a potential mediator of microbiota perturbations during WT enteric salmonellosis. These studies offer a new model of S. typhimurium-induced intestinal disease that retains the three participants of the disease process and further defines the role of virulence factors, the host microbiota, and inflammation in S. typhimurium-induced intestinal disease.

Introduction

It is now appreciated that an enteric infectious process is a complex tripartite relationship between the invading pathogen, the host and the host's indigenous microbiota.¹ Furthermore, pathogen-elicited host responses can prove detrimental to intestinal microbiota, promoting pathogen colonization.^2,3 However, it is not known which aspects of the inflammatory response are responsible for the observed effects on microbiota, whether the indigenous microbial population is similarly affected by all inflammatory mediators, and which virulence factors are responsible for the observed effects.

Salmonella enterica serovar Typhimurium (S. typhimurium) is a gastrointestinal pathogen and a frequent cause of food poisoning worldwide, accounting for much morbidity and mortality. S. typhimurium's pathogenesis is dependent on virulence factors encoded on a number of pathogenicity islands, termed Salmonella pathogenicity islands, or SPI.⁴ Two type III secretion systems (T3SS) encoded on SPI-1 and SPI-2 mediate many aspects of both the intestinal and systemic phases of S. typhimurium infection, including invasion of the intestinal epithelium,^5,6 induction of the inflammatory response,⁷ and intracellular survival.⁸ The SPI-1-encoded T3SS and associated effector proteins are crucial for S. typhimurium invasion of the intestinal epithelium and also play an important role in the induction of the host inflammatory response. Mutation of invA, a component of the SPI-1 apparatus (invAmut strain), abrogates the ability of S. typhimurium to invade cultured intestinal epithelial cells.⁹ The SPI-2-encoded T3SS is active during the intracellular stage of S. typhimurium infection,⁶ allowing S. typhimurium to evade killing in the phagocytic vacuole and instead establish a successful niche for intracellular survival. Deletion of ssaR, a component of the SPI-2 apparatus (ssaRmut strain), renders SPI-2 non-functional and S. typhimurium susceptible to intracellular killing.

Careful dissection of mechanisms accounting for S. typhimurium-induced intestinal disease can further our understanding of the disease process and facilitate development of better prevention and treatment methods. Until recently, no convenient animal model system was available to study the intestinal phase of the S. typhimurium-induced disease. The murine host is normally only susceptible to the systemic illness (reminiscent of S. typhi-induced enteric fever in humans), while the bovine infectious colitis model is genetically far more heterogeneous and clearly not as amenable to manipulation, complicating data interpretation. In the 1950s and 60s it was noted that treatment of mice with a high dose of an antibiotic renders them susceptible to S. typhimurium-induced intestinal disease.^10,11 In 2003, Hardt's laboratory capitalized on these findings to further develop a murine S. typhimurium colitis model, which allows analysis of pathogen and host factors involved in the intestinal disease process.¹² However, this model requires pre-treatment of mice with massive doses of streptomycin (20 mg/mouse), which severely reduces the total numbers of the murine intestinal microbiota.² While this model (high dose or HD model) offers great advantages to the study of the host and the pathogen, the role of the third participant in the process—the intestinal microbiota—cannot be adequately assessed.

We present here a low dose streptomycin pre-treatment murine model of S. typhimurium-induced cecal inflammation in which total microbial levels in uninfected mice remain unaltered.¹³ This facilitates the examination of microbiota perturbations during infection. Using this model we examined which SPIs of S. typhimurium induce microbiota alterations and which host response component promotes them. We found that while both WT and invAmut S. typhimurium strains induce significant inflammation in the mouse ceca, only infection with the WT strain caused a pronounced reduction in total microbial numbers. This microbial depletion was associated with a large neutrophil infiltrate in the cecal lumina of WT-infected mice. However infection with invAmut and ssaRmut strains did not yield this phenotype. Neutrophils are thus implicated as one of the host agents involved in the inflammation-induced damage to the host microbiota. We also demonstrated that SPI-1-encoded virulence factors are necessary for S. typhimurium to induce a host response that adversely affects the indigenous microbial population.

Results

SPI-2 is necessary to maintain S. typhimurium cecal colonization in LD model of murine enterocolitis.

For the LD model, C57BL/6 mice were pre-treated with a low dose of streptomycin (450 mg/L in drinking water for two days) and then infected with one of three S. typhimurium strains. We examined colonization of the infected mouse ceca by WT, invAmut and ssaRmut S. typhimurium over the duration of infection to assess their ability to colonize in the presence of adequate numbers of the intestinal microbiota. S. typhimurium colonization of murine ceca in the HD model (in which the mice are pre-treated with 20 mg streptomycin 24 hrs prior to infection) is included for comparison. In the LD model, colonization by all three strains reached maximal levels as early as 1 day post infection (p.i.), peaking at 10⁹–10¹⁰ cfu/g cecum (Fig. 1). WT and invAmut S. typhimurium colonization remained relatively unaltered for the entire duration of the infection (Fig. 1 and Table 1). Cecal colonization by ssaRmut S. typhimurium declined gradually over the time course of infection with significant reductions in Salmonella burden at days 5 and 6 p.i., compared to earlier time-points (Fig. 1). Moreover, ssaRmut S. typhimurium cecal burden was significantly lower than those of the WT strain at days 4 and 5 p.i. and lower than those of the invAmut strain at days 5 and 6 p.i. (Table 1). While these colonization dynamics are similar to those previously described in the HD model,^14,15 they differed from those obtained by us in replicating the HD infection model for comparison purposes (Fig. 1). In our results for the HD colonization dynamics, ssaRmut was able to colonize the infected mice at the same level for the whole duration of the infection, similar to WT and invAmut strains. Colonization by all three strains was somewhat lower in the HD model, compared to the LD model, likely due to the lower infectious dose utilized in the HD model (see Materials and Methods for explanation).

Figure 1.

Comparison of S. typhimurium colonization dynamics in the LD and HD murine infection models. LD model: Mice were treated with 450 mg/L of streptomycin for twp days in drinking water. After antibiotic withdrawal, mice were infected with 2.7 × 10⁸ cfu of indicated S. typhimurium strain; two to eight mice per group. HD model: Mice were treated with 20 mg streptomycin 24 hours prior to infection with 2 × 10⁵ cfu of indicated S. typhimurium strain; two to four mice per group. S. typhimurium colonization was enumerated by plating serial dilutions of cecum homogenates on XLD plates with 100 µg/ml streptomycin. N.I. = uninfected group; D = days post infection.

Table 1.

Statistical analysis of cecum colonization by WT, invAmut and ssaRmut S. typhimurium strains over time course of infection in LD model of enterocolitis

	Day 1		Day 3		Day 4		Day 5		Day 6
vs.	invAmut	ssaRmut	invAmut	ssaRmut	invAmut	ssaRmut	invAmut	ssaRmut	invAmut	ssaRmut
WT	ns	ns	ns	ns	ns	p < 0.05	ns	p < 0.01	NA	NA
invAmut	-	ns	-	ns	-	ns	-	p < 0.001	-	p < 0.001

ns, not significant; NA, not available. Colonization by different strains at the same time points was compared. All p-values were calculated using ANOVA, except for day 6 ssaRmut vs. invAmut comparison, which was calculated using unpaired t-test, as only two groups of mice survived to this time point.

In the LD model, mice infected with WT and invAmut S. typhimurium quickly became very moribund with no surviving WT-infected mice past day 5 and only 2 surviving invAmut-infected mice at day 6 p.i.. Mice infected with ssaRmut S. typhimurium appeared considerably less sick (with less ruffling of the fur, absence of hunched posturing and appearing subjectively less cachectic), with all the infected mice surviving at day 6 p.i.. These differences are most likely due to the inability of the ssaRmut strain to cause lethal systemic illness.¹⁶ In agreement with that, WT and invAmut S. typhimurium were able to colonize the spleens of infected mice to significantly higher levels than ssaRmut S. typhimurium (data not shown).

PI-2 is necessary for induction of cecal inflammation in LD model of murine enterocolitis.

Next we examined the intestinal inflammation caused by the three selected S. typhimurium strains in this infection model. As S. typhimurium induces the most pronounced inflammatory changes in the ceca of the infected mice,^12,14 the presence and extent of cecal inflammation was evaluated as a marker of intestinal inflammation. Similar to the situation in the HD model,¹² treatment of uninfected mice with a low dose of streptomycin produced a notable enlargement of the ceca. However, this enlargement failed to reach statistical significance (data not shown). Infection-associated inflammatory changes resulted in the histopathological manifestations discussed below and acted to reverse the antibiotic-induced cecal enlargement (data not shown).

In the LD model, infection with WT S. typhimurium caused significant cecal inflammation in infected mice as early as day 3 p.i. (Fig. 2B). Inflammation was characterized by increased amounts of inflammatory infiltrate, starting from the submucosa and spreading to the lumen. Increased submucosal and mucosal thickness, mounting epithelial disorganization, presence of mucinous plugs in the crypts, and dead epithelial cells in the lumina of the tissues also indicated increased pathology (Fig. 2A). The inflammatory changes persisted until day 5 p.i., at which time the mice became too moribund and were sacrificed (Fig. 2). WT S. typhimurium-infected mice in the HD model demonstrated similar histopathological changes in their lumina as a result of infection (Fig. 2A), with the extent of pathology reaching statistical significance as early as day 1 post-infection, and remaining high until the end of the infection time-course (Fig. 2B).

Figure 2.

Comparison of S. typhimurium-induced intestinal inflammation in the LD and HD murine infection models. LD model: Mice were treated with 450 mg/L streptomycin in drinking water for two days. Following antibiotic withdrawal, mice were infected with 2.7 × 10⁸ cfu of the specified S. typhimurium strain. Two to 8 mice per group were used. HD model: Mice were treated with 20 mg streptomycin 24 hours prior to infection with 2 × 10⁵ cfu of indicated S. typhimurium strain; 2 to 4 mice per group. p values were calculated using Kruskal-Wallis with Dunn's post test with 95% confidence interval. N.I. = uninfected group, D = days post infection. (A) Tissues were harvested, fixed in formalin and stained with H&E. In the LD model, WT and invAmut S. typhimurium-infected sections show escalating pathology over time. In the HD model, WT-infected mice show severe pathology throughout the duration of infection. The degree of pathology is gradually increasing in the invAmut-infected mice, while pathological changes in ssaRmut-infected mice peak early in infection, with following gradual reduction in pathological indices. In both infection models pathological changes are indicated by rising levels of inflammatory infiltrate starting from submucosa and spreading to the lumen, as well as evident crypt abscesses; increasing epithelial disorganization is indicated by mucinous plugs in crypts, mounting epithelial regenerative changes, desquamation and presence of dead epithelial cells in the lumen. Arrow-heads indicate lumen. Scale bar (in N.I. micrograph) is 100 µm. (B) Quantification of indicators of pathology. Groups marked with an asterisk are significantly different from uninfected group, p < 0.05 or smaller.

In the LD model, mice infected with the invAmut strain did not have any notable inflammatory changes in their ceca at the early time points post-infection (Fig. 2), consistent with previous findings in the HD model^14,17 and also similar to our HD comparison experiment (Fig. 2). At the late time points of infection, the ceca became appreciably inflamed, reaching significant levels at day 5 p.i. (Fig. 2B), also in agreement with previous findings regarding the role of SPI-2 in intestinal inflammation.^14,15 Cecal inflammation at day 6 p.i. was similar to that at day 5 p.i. but did not reach significance when compared to uninfected mice. This is probably due to only 2 mice surviving at this time point. Of note is the fact that while in the LD model cecal histopathology in the invAmut infected mice only became appreciable at day 5 p.i., in the HD model inflammatory changes were notable as early as day 3 p.i., reaching statistically significant level by day 4 p.i. (Fig. 2).

In the LD model, cecal inflammation induced by the invAmut was similar to that induced by WT S. typhimurium, with epithelial degeneration, increased submucosal and mucosal thickness, presence of cellular debris in the lumen, mucinous plugs and crypt abcesses. Although the invAmut-induced cecal inflammation was pronounced and significant, it never reached the extent of that induced by the WT strain, remaining significantly lower at the same time points for the duration of the infection time course (p = 0.004 at day 5 p.i., Mann-Whitney U-test). This is in contrast to the situation in the HD model.¹⁵

In the LD model, mice infected with ssaRmut S. typhimurium did not exhibit any significant cecal inflammation for the duration of the infection and cecal tissues were morphologically similar to those of the uninfected mice (Fig. 2). Thus in the presence of adequate numbers of the intestinal microbiota the SPI-1 pathogenicity island alone is not sufficient to trigger cecal inflammation. This is in contrast to the previously described findings in the HD model, in which significant (albeit significantly lower than WT-induced) cecal inflammation was observed at 2 days p.i. in mice infected with a SPI-2 mutant S. typhimurium strain.^14,15 In our HD model comparison studies, inflammatory changes induced by the ssaRmut strain were the greatest at day 1 p.i., thereafter gradually declining below statistical significance, although still remaining appreciable (Fig. 2).

Both SPI-1 and SPI-2 are necessary to produce a reduction in the total numbers of the host intestinal microbiota in LD model of murine enterocolitis.

To determine the effect of infection with the three S. typhimurium strains on the host intestinal microbiota, we evaluated the total bacterial numbers, the proportion of γ-Proteobacteria and microbial community profiles in the murine ceca over time. We expected to see profound perturbations to the host microbiota during infection with WT and invAmut S. typhimurium due to the intestinal inflammation observed in response to these two strains (Fig. 2), as previous microbiota studies in infectious colitis linked development of intestinal inflammation to microbiota perturbations.^2,3,18

Infection with WT S. typhimurium resulted in a significant decrease in total cecal bacterial numbers, as assessed by SYBR green staining, starting at day 3 p.i. (Fig. 3A). The observed reduction was consistent among animals and the intestinal microbiota numbers did not recover for the duration of the experiment (Fig. 3A). Total bacterial numbers in mice infected with either the invAmut or the ssaRmut S. typhimurium strains were unaffected by the infection, except for a transient reduction observed at day 4 in invAmut-infected mice (Fig. 3A). This reduction, however, did not persist, unlike the situation during WT infection.

Figure 3.

Intestinal microbiota is extensively modified during infection with WT, but not invAmut or ssaRmut S. typhimurium strains in the LD model of enterocolitis. Mice were treated with 450 mg/L of streptomycin for two days in drinking water. After antibiotic withdrawal, mice were infected with 2.7 × 10⁸ cfu of the indicated S. typhimurium strains. Two to 8 mice per group were used. p values were calculated using one-way ANOVA with Bonferroni post test with 95% confidence interval. Time points marked with an asterisk are significantly different from control mice, p < 0.05 or smaller. N.I. = uninfected group; D = days post infection. (A) Total bacteria were enumerated by SYBR green staining. (B) Proportion of γ-Proteobacteria was determined by FISH.

To assess the perturbations in the composition of the intestinal microbiota as a result of infection, the proportion of cecal γ-Proteobacteria was determined using fluorescent in-situ hybridization (FISH) and the cecal microbial community profiles were assessed via terminal restriction lengths polymorphism (tRFLP) analysis.

Infection with WT S. typhimurium resulted in a prominent and significant increase in the proportion of cecal γ-Proteobacteria starting at day 3 p.i. (Fig. 3B). The majority of these γ-Proteobacteria were the infecting Salmonella (Suppl. materials, Fig. S1). In contrast, infection with each of the mutant S. typhimurium strains did not result in a pronounced or consistently significant increase in the proportion of cecal γ-Proteobacteria (Fig. 3B), most likely due to the fact that the total commensal bacterial numbers were not reduced and consequently the infecting Salmonellae did not account for a large proportion of the total microbial population. The observed increases did not last throughout the time course of the infection and were far smaller than the increases observed during infection with the WT strain.

When cecal microbial community profiles were assessed, it was noted that the bacterial communities of WT-infected mice tend to cluster together and away from the bacterial communities of invAmut- and ssaRmut-infected mice (Fig. 4). This was true for days 3, 4, and 5 p.i. Cecal microbial community profiles at day 1 p.i. were not analyzed, since no notable changes were observed at this time point in either total bacterial numbers or proportion of γ-Proteobacteria (Fig. 3). Microbial communities of invAmutand ssaRmut-infected mice overlapped with each other and with uninfected mice at days 3 and 5 p.i. (Fig. 4). At days 4 and 6 p.i. there was less of an overlap between the bacterial communities of the mutant strains-infected mice, and at day 6 p.i. they also separated more from the communities of the uninfected mice (Fig. 4), likely due to more extensive alterations to the composition of the microbial community later in the infectious process.

Figure 4.

Microbial community profiles of WT S. typhimurium-infected mice cluster away from those of invAmut- and ssaRmut-infected mice in the LD model of enterocolitis. Mice were treated with 450 mg/L streptomycin in drinking water for two days. Following antibiotic withdrawal, mice were infected with 2.7 × 10⁸ cfu of the specified S. typhimurium strain. Two to 8 mice per group were used. Ceca were collected at days 3–6 post infection. Bacterial 16s rRNA was amplified for TRFLP analysis as described in Materials and Methods. NMS plots show separation of treatment groups based on changes in microbial composition as assessed by T-RFLP.

All methods of microbiota analysis demonstrated that infection with the WT S. typhimurium strain results in significant alterations to both the numbers and composition of the intestinal microbiota. Infection with either of the S. typhimurium mutant strains produced alterations to the microbial community composition, as was evidenced by the tRFLP analysis, but failed to significantly and consistently reduce the total numbers of the intestinal microbes. Furthermore, the community profiles of the mutant-infected mice were more similar to each other and at some time points to those of the uninfected mice, than they were to the community profiles of the WT-infected mice. Therefore, although invAmut S. typhimurium was able to provoke an inflammatory response in the infected host, the induced inflammation did not have the same effect on host intestinal microbiota (both in terms of numbers and composition) as inflammation caused by WT S. typhimurium infection.

Both SPI-1 and SPI-2 are necessary for neutrophil infiltration into Salmonella-infected ceca in LD model of murine enterocolitis.

Although both the WT and the invAmut S. typhimurium strains were able to cause cecal inflammation in the infected mice, only infection with the WT strain also significantly reduced total intestinal microbial numbers. Possible explanations accounting for this disparity are differences in the extent of the induced inflammation, the type of the induced inflammatory response, or direct interactions between SPI-1 effector(s) and the indigenous microbiota. As the indigenous microbiota is complex and the vast majority of its constituents cannot yet be cultured, it was not feasible to evaluate the latter option. However, we did assess the first two possible explanations.

Although the extent of the inflammation induced by the invAmut was less than that induced by the WT strain for each given time point, the pathology score for invAmut at day 5 p.i. was not significantly lower than that of the WT strain at day 3 p.i. (p > 0.05, Mann-Whitney U test) (Fig. 2B). However, no profound changes to the microbiota were observed in the invAmut-infected mice at day 5 p.i., unlike the situation in WT-infected mice at day 3 p.i. (Fig. 3). Consequently, the differences in the type of the inflammatory response induced by the two strains are more likely to be accountable for the disparity in their effect on the host intestinal microbiota.

S. typhimurium infection is known to induce a strong inflammatory cellular response,¹⁹ with many types of inflammatory cells present in the inflamed murine intestinal tract.^14,20,21 Consequently, we hypothesized that the presence of different types of inflammatory cells during infection with the three examined S. typhimurium strains might account for the observed differences in the effect of infection-induced colitis on the host intestinal microbiota. When we examined the presence of T-cells (CD3 positive) and granulocytes (GR1 positive) in the infected mouse ceca, we did not observe striking differences between the WT- and invAmut-infected mice (data not shown). However, the situation was quite different when neutrophil infiltration was examined.

Immunostaining of cecal necropsies for myeloperoxidase (MPO) granules found in neutrophils showed a neutrophil influx into the lumina of the WT-infected mice starting as early as day 3 p.i. (Suppl. materials, Fig. S2) and persisting until day 5 (Fig. 5), at which point all WT-infected mice were sacrificed. The presence of neutrophils correlated with the increase in pathological indices and microbiota perturbations observed during WT S. typhimurium infection (Figs. 2–4, Suppl. materials Fig. S1). The presence of luminal neutrophils was first observed in WT-infected tissues at day 3 p.i. (Suppl. materials, Fig. S2), coincidental with appearance of a significant increase in pathology in the infected ceca (Fig. 2) and significant microbiota alterations (Fig. 3). Increasing presence of luminal neutrophils later in infection (Fig. 5 and Suppl. Materials Fig. S2) was accompanied by augmenting pathology (Fig. 2) and greater microbiota alterations (Fig. 3). In contrast, no positive neutrophil staining was observed in the lumina of the mice infected with either invAmut or ssaRmut S. typhimurium strains (Fig. 5 and Suppl. Materials Fig. S2). This conspicuous absence of luminal neutrophils went along with the absence of profound microbiota perturbations during infection with these two mutant strains (Figs. 3 and 4 and Suppl. Materials Fig. S1).

Figure 5.

Murine infection with WT, but not invAmut and ssaRmut S. typhimurium results in cecal infiltration of neutrophils in the LD model of enterocolitis. Mice were treated with 450 mg/L streptomycin in drinking water for two days. Following antibiotic withdrawal, mice were infected with 2.7 × 10⁸ cfu of the specified S. typhimurium strain for five days. Four to 8 mice per group were used. Images are from a representative mouse from each group. Formalin-fixed, paraffin embedded cecum sections were stained for nuclear DNA (with DAPI [4′,6′-diamidino-2-phenylindole]; blue in overlay), neutrophil myeloperoxidase (green in overlay), and Salmonella LPS (red in overlay). Ceca of mice infected with WT S. typhimurium exhibit positive luminal staining for neutrophils (granular staining surrounding nuclei in the lumen). Lumina of murine ceca infected with each of the mutant strains did not stain positive for neutrophils. All of the infected tissues stained positive for Salmonella. “L” indicates lumen in overlay images; arrows indicate positive staining of myeloperoxidase granules. Images are pseudocolour, scale bar (overlay ssaRmut panel) is 200 µm.

In summary, we have shown that both functional SPI-1 and SPI-2 are required to effect neutrophil recruitment into the lumina of the infected mouse ceca. Additionally, an association was demonstrated between the presence of luminal neutrophils and infection-induced microbiota depletion.

Discussion

As the role of the resident intestinal microbiota in the progression of enteric infections gains increasing attention, convenient models for dissecting the particulars of the host-pathogen-microbiota interactions are needed. Using a low dose streptomycin pre-treatment model (LD model) of murine salmonellosis that retains unaltered numbers of the host microbiota¹³ while supporting the establishment of WT S. typhimurium-induced intestinal inflammation, we investigated the contribution of SPI-1 and SPI-2 to the induction of murine cecal inflammation and microbiota perturbations over 6 days of infection by S. typhimurium. While the previously described HD model of murine intestinal salmonellosis¹² provided a very useful tool for the in vivo assessment of the contribution of S. typhimurium virulence factors and host defenses to the progression of intestinal infection, the very large dose of streptomycin used to pre-treat the mice in this model greatly decreased the numbers of the resident microbiota,² thus precluding a fair evaluation of its role in the infection progression. Use of the LD model presented here allows a better assessment of the role of the microbiota in the infection progression, as well as of the impact of S. typhimurium on the microbiota.

The majority of the studies in the HD model focused on the role of either the bacterial or host factors in the progression of the murine intestinal salmonellosis and did not examine the effect of the infection on the microbiota. Some studies focusing on microbiota analysis in the HD model indicated that pretreatment with a high dose of streptomycin significantly alters the pre-infection microbiota composition of both the ileum²² and the cecum,^2,22 promoting enhanced colonization by S. typhimurium and greater pathology. Previous studies in murine models of S. typhimurium infection involving antibiotic pre-treatment have demonstrated that both major and minor perturbations in the intestinal microbiome predispose the host to more severe infections.^13,23

Previously published results on the effect of enteric infections on the microbiota have indicated that infection-associated inflammation acts to extensively modify and reduce indigenous microbes.^2,3,18 We sought to evaluate the effect of infection with the three S. typhimurium strains on the total numbers and composition of the murine intestinal microbiota. We found that while inflammatory changes associated with WT infection, induced extensive perturbations in the murine microbiota (as previously described in both the LD¹³ and the HD models²), inflammation induced during infection with invAmut strain failed to affect the host microbiota in the same way (Figs. 3 and 4 and Suppl. Materials Fig. S131 y). A study in a FvB mouse strain with no antibiotic pre-treatment indicated that no inflammation was associated with a 3 day long infection with either a SPI-1 or a SPI-2 S. typhimurium mutant, while there occurred partial changes to the microbiota following infection with a SPI-2 mutant.²⁴ The reported partial changes were in the proportion of some of the microbiota species and no alterations to the total numbers of the microbiota were observed, while infection with a WT S. typhimurium strain caused a 95% decrease in microbiota at day 7 p.i..²⁴

S. typhimurium infection is known to induce a strong inflammatory cellular response,¹⁹ with many inflammatory cells present in the inflamed murine intestinal tract.^14,20,21 Neutrophils have been observed in high numbers in the intestinal tissues of human patients and have been shown to dominate the stool leukocyte population.²⁵ Significant microbiota depletion during the inflammatory response in mice has been shown to occur during Citrobacter rodentium and S. typhimurium infections, and intestinal inflammation elicited by both these pathogens is characterized by neutrophil infiltration.^25,26 Neutrophil infiltration into the colonic mucosa is also the hallmark of inflammatory bowel diseases (IBD)²⁷ and is a prominent feature of dextran sodium sulfate (DSS)—induced colitis used to model IBD in laboratory animals.²⁸ A decrease in the microbial population was observed both in IBD patients²⁹ and in murine DSS-induced colitis.³ Coburn and colleagues have shown that neutrophils are differentially recruited by the WT, invAmut and ssaRmut S. typhimurium strains into the ceca of mice in HD model.¹⁴ This prompted us to hypothesize that variations in neutrophil recruitment by the different S. typhimurium strains might also occur in this model system, potentially accounting for the observed differences in the effects on the intestinal microbiota. Indeed, when neutrophil infiltration into the lumina of the diseased ceca was evaluated, it was observed that while WT infection resulted in a massive neutrophil influx starting at day 3 p.i. and persisting until the end of the infection time-course, no luminal neutrophils were present in the mice infected with the invAmut, or the ssaRmut strains (Fig. 5 and Suppl. Materials Fig. S2).

Neutrophils harbour a diverse microbicidal arsenal, including microbicidal granules, phagocytic abilities, production of reactive oxygen species (ROS)³⁰ and neutrophil extracellular traps (NETs).³¹ Members of the host intestinal microbiota are likely to be more susceptible to these microbicidal mechanisms than the invading pathogens, such as S. typhimurium, that have evolved to resist host defense mechanisms. The presence of ROS is likely to be particularly detrimental to the strictly anaerobic bacteria that make up a large proportion of the intestinal microbiota. Consequently S. typhimurium-induced neutrophil influx could account for the preferential elimination of the anaerobic bacteria from the CFB phylum observed during WT S. typhimurium infection.^2,13 Failure of the SPI-1 mutant to induce neutrophil influx in the infected mice could explain the absence of severe perturbations to the intestinal microbiota despite the presence of significant inflammation. In this manner neutrophil recruitment appears to be directly beneficial to S. typhimurium, illuminating additional facets of the question “who drives the neutrophil influx during enteric salmonellosis?“.²⁵

Upon assessment of the intestinal pathology induced in our model, some differences were noted compared to the previously described HD model. Although previous in vivo work in the HD model indicates that early conspicuous murine cecal inflammation (at day 2 post infection) can be induced by SPI-1 alone,^14,15,17 we now show that in the presence of adequate numbers of the intestinal microbiota, SPI-1 is insufficient to induce significant inflammatory changes in the murine cecum (Fig. 2).

The presence of SPI-2 alone was also shown to be insufficient to induce intestinal inflammation early in infection, similar to the situation in HD model, while notable inflammation developed later in infection. A similar contribution of SPI-2 to later stages of the intestinal inflammation was also noted in the HD model in both previous and current studies,^14,15 while others have demonstrated that both SPI-1 and SPI-2 alone are able to induce significant cecal inflammation at the later time points of infection.³²

The WT strain was able to cause significant cecal inflammation in the LD model starting at day 3 p.i. In previous reports of the HD model, WT S. typhimurium-induced cecal inflammation reached its maximum intensity by 48 hrs p.i., persisting at the same level until day 5 p.i.^14,15 however, we found significant inflammatory changes as early as day 1 p.i. with a peak at day 4 p.i.. In contrast, we have found that in the presence of normal numbers of host microbiota (the LD model), inflammation takes a longer time to develop, reaching significantly high intensity only at day 3 p.i., and continuing to increase until 4 days p.i., at which point it peaks.

Presented here is a murine model of enteric salmonellosis that provides a tool for elucidating the details of the host-pathogenmicrobiota interactions during S. typhimurium infection. This model has increased our understanding of the relative roles of SPI-1 and SPI-2 in the generation of intestinal inflammation and points to their contributions to the profound perturbations in the host intestinal microbiota observed during S. typhimurium infection. Our results indicate that while inflammation induced during infection with WT S. typhimurium produces severe disturbances in the host intestinal microbiota, inflammatory changes associated with an invAmut infection fail to generate the same effect. WT S. typhimurium-induced neutrophil infiltration was implicated as one of the potential mediators of microbiota perturbations.

Materials and Methods

Bacterial strains.

Salmonella typhimurium SL1344 (WT),³³ invA mutant (invA::kan SB103, defective in a structural gene of the SPI-1 type III apparatus; ‘invAmut’)³⁴ and ssaR mutant (ΔssaR, in frame deletion of a structural gene of the SPI-2 type III apparatus; ‘ssaRmut’)³⁵ were grown overnight (ON) shaking (200 rpm) in Luria-Bertani broth (LB) with 100 ug/ml streptomycin over night.

Mice.

Inbred C57BL/6 female mice (Jackson Laboratory, Bar Harbor, Maine, USA) were housed in the animal facility at the University of British Columbia in direct accordance with guidelines drafted by the University of British Columbia's Animal Care Committee and the Canadian Council on the Use of Laboratory Animals and infected at 4.5–5.5 weeks of age. Mice were fed a standard sterile chow diet (Laboratory Rodent Diet 5001, Purina Mills, St. Louis, Missouri) ad libitum throughout the experiments.

Mouse infections, tissue collection and S. typhimurium enumeration.

LD infections. Mice were treated with 450 mg/L of streptomycin (Sigma) in drinking water as described previously.¹³ As mice drink, on average, 3 ml of liquid per day,³⁶ the average consumed dose of streptomycin was 1.35 mg/mouse/day. Control mice were given sterilized non acidified drinking water without the antibiotic. After 2 days the antibiotic was withdrawn and mice were infected with the appropriate strain of S. typhimurium at 2.7 × 10⁸ cfu/mouse by oral gavage. Uninfected control mice were given 100 µl of sterile LB broth. At an indicated time point post-infection the mice were euthanized by CO₂ asphyxiation and tissues were harvested aseptically for further evaluation.

HD infections. Mice were fasted for 4 hours and then 20 mg streptomycin (Sigma) was administered by oral gavage to each mouse. 24 hours later mice were again fasted for 4 hours, following which they were infected with the appropriate strain of S. typhimurium at 2 × 10⁵ cfu/mouse by oral gavage. A lower inoculum was chosen than for LD infections, as experience in our laboratory indicates that at higher infective doses mice are unlikely to survive beyond 3–4 days of infection, and we were planning an infection time course of 5 days.

Ceca were collected in 1 ml of sterile PBS on ice and homogenized with a MixerMill 301 (Retsch, Newtown, PA, USA). Serial dilutions of the homogenates were plated on Xylose Lysine Deoxycholate (XLD) (Oxoid) or LB agar plates containing 100 µg/ml streptomycin to enumerate S. typhimurium colonization.

Histopathology.

Cecal tips were fixed in 10% neutral buffered formalin overnight and then placed into 75% ethanol. Fixed tissues were embedded in paraffin and cut into 5 µm sections by Wax-it Histology Services (Vancouver, BC, Canada). Tissues were stained with hematoxylin and eosin (H&E) using standard techniques by Wax-it Histology Services. Pathological scores were assigned as previously described.¹⁴

Pathology sections were viewed and images were taken using Zeiss Axioskop 2 microscope.

Immunofluorescence.

Five-micrometre sections of fixed cecal tips were deparaffinized and re-hydrated. Following rehydration tissues were subjected to heat-induced antigen retrieval in citrate buffer. Immunostaining was carried out using antibodies against MPO (Thermo Scientific, Fremont, CA) to visualize neutrophils and anti-Salmonella LPS (BioDesign, Saco, ME) to visualize S. typhimurium. Secondary antibodies conjugated to Alexa fluorophores were used.

Immunofluorescence sections were viewed and images were taken using an Olympus 1X81 microscope.

Microbiota assessment.

SYBR green staining. SYBR Green (Invitrogen) staining, which binds to dsDNA, was used to enumerate the total numbers of microbiota in the samples, as described previously.¹³ Briefly, 2 to 40 µl of samples were stained with 0.25 µl SYBR Green (Invitrogen) and viewed with an Olympus 1X81 microscope. Three fields were randomly chosen; the number of cells were counted and averaged. The counts were made in a microscope field of a known diameter and corrected to the volume of sample used.

Fluorescent in situ hybridization (FISH).

FISH was used to determine the proportion of γ-Proteobacteria in the samples. General EUB338 probe (5′ GCT GCC TCC CGT AGG AGT 3′)³⁷ and GAM42a probe (5′ GCC TTC CCA CAT CGT TT 3′)³⁸ were used and the protocol described previously was followed.³

Terminal restriction fragment polymorphism (tRFLP) analysis.

tRFLP was used to evaluate microbiota community profiles. Genomic DNA was isolated from murine ceca using QIAamp DNA stool minikit (Qiagen, Maryland, USA) according to manufactures instructions with the addition of a bead beating step. Bacterial 16S rRNA was amplified with primer 8F (FAM labeled) (5′-AGA GTT TGA TCM TGG CTC AG-3′) and 926R (5′-CCG TCA ATT CCT TTR AGT TT'-3′). Each 25 µL reaction mixture contained 1x PCR buffer, 2 mM MgCl₂, 0.4 mM of each of the deoxynucleoside triphosphate, 0.4 µM forward and reverse primers, 1.25 U of Taq polymerase (Fisher Scientific), 0.1 mg/ml bovine serum albumin and approximately 50 ng genomic DNA. Cycling conditions were: 5 min of denaturation at 94°C, 30 cycles of 0.5 min at 94°C, 0.5 min at 55°C, 1 min at 72°C, and a final 7 min extension step at 72°C. PCR products were digested with MspI (New England Biosciences) and 1× Neb Buffer for 3 hours at 37°C. Samples were diluted 1:10 and sent to the Nucleic Acid Protein Service Unit (University of British Columbia) for analysis on a 3730 DNA Analyser with a GeneScan 1200 LIZ size standard (Applied Biosystems). The profiles were imported into the GeneMarker v1.75 software (SoftGenetics, State College, PA) and fragment lengths were determined using the internal size standard and local Southern algorithm. Similarities of the microbial community structures between groups were analyzed by NMS analysis of the normalized data set using PC-ORD 5.0 software.

Statistical analysis.

One-way ANOVA with Bonferroni post test or Kruskal-Wallis with Dunn's post test was performed using a 95% confidence interval. Where stated, unpaired t-test or Mann-Whitney U-test was used using a 95% confidence interval. All analyses were performed using GraphPad Prism version 4.0. Differences were considered to be significant with p < 0.05 or smaller.