Dietary choice affects Shiga toxin-producing Escherichia coli (STEC) O157:H7 colonization and disease

Significance

We demonstrated that dietary fiber content affects susceptibility to Shiga toxin (Stx)-producing Escherichia coli (STEC) infection in mice. We showed that high fiber diet (HFD)-fed mice had elevated levels of butyrate, a beneficial gut metabolite that paradoxically enhances the cell-killing capacity of Stx. We also found that the amount of gut bacteria in HFD-fed mice increased whereas the percent of commensal Escherichia species (spp) decreased compared with animals fed a low fiber diet (LFD). These changes led to higher E. coli O157:H7 colonization levels, more weight loss, and greater rates of death in HFD-fed than in LFD-fed STEC-infected animals.

Abstract

The likelihood that a single individual infected with the Shiga toxin (Stx)-producing, food-borne pathogen Escherichia coli O157:H7 will develop a life-threatening sequela called the hemolytic uremic syndrome is unpredictable. We reasoned that conditions that enhance Stx binding and uptake within the gut after E. coli O157:H7 infection should result in greater disease severity. Because the receptor for Stx, globotriaosylceramide, is up-regulated in the presence of butyrate in vitro, we asked whether a high fiber diet (HFD) that reportedly enhances butyrate production by normal gut flora can influence the outcome of an E. coli O157 infection in mice. To address that question, groups of BALB/c mice were fed high (10%) or low (2%) fiber diets and infected with E. coli O157:H7 strain 86-24 (Stx2+). Mice fed an HFD exhibited a 10- to 100-fold increase in colonization, lost 15% more body weight, exhibited signs of morbidity, and had 25% greater mortality relative to the low fiber diet (LFD)-fed group. Additionally, sections of intestinal tissue from HFD-fed mice bound more Stx1 and expressed more globotriaosylceramide than did such sections from LFD-fed mice. Furthermore, the gut microbiota of HFD-fed mice compared with LFD-fed mice contained reduced levels of native Escherichia species, organisms that might protect the gut from colonization by incoming E. coli O157:H7. Taken together, these results suggest that susceptibility to infection and subsequent disease after ingestion of E. coli O157:H7 may depend, at least in part, on individual diet and/or the capacity of the commensal flora to produce butyrate.

Shiga toxin (Stx)-producing Escherichia coli (STEC) infections continue to be a significant health burden in the United States. There are typically 15–20 outbreaks of STEC in the United States per year (1) that result in 265,000 illnesses, 3,600 hospitalizations, and 30 deaths annually (2). The United States Department of Agriculture estimates a total cost of about $500 million per year in health-related costs in the United States due to STEC infection (3).

STEC, such as the most frequently isolated serotype in the United States, E. coli O157:H7 (1), are gut commensals of cattle. However, in humans, E. coli O157:H7 can cause diarrhea and hemorrhagic colitis after ingestion of as few as <50–300 organisms (4, 5). Such infections typically occur after individuals consume E. coli O157:H7-contaminated beef, fresh vegetables, water, or unpasteurized juice (6). On occasion, E. coli O157:H7 infections result from person-to-person spread of this low-infectious dose pathogen (6). The usual disease progression is as follows: 3 d after the ingestion of contaminated food or water, diarrhea, abdominal pain, and vomiting begin; frank, bloody diarrhea follows 2–3 d later. The bloody diarrhea generally resolves after 4 or 5 d. However, in 4–30% (7–10) of patients, the potentially life-threatening, hemolytic uremic syndrome (HUS) develops (6). HUS is characterized by acute kidney malfunction, microangiopathic hemolytic anemia, and thrombocytopenia (11).

One aspect of STEC pathogenesis that is not clear is why a significant disparity in age and sometimes sex exists among those with STEC-related HUS. Children <10 y old are 10 times more likely to develop HUS following infection with STEC (12). In two recent large outbreaks, women developed HUS in disproportionate numbers (9, 13). Studies that address why children are more likely than adults to develop HUS suggest that a difference in complement activation (14, 15), platelet activation (16), or in nitric oxide production (17) may explain the disparity.

The Stx produced by STEC is the critical virulence factor that not only may contribute to intestinal epithelia damage (18, 19) but is required for the development of HUS (20). STEC may produce Stx1, Stx2, or both toxins. Stx1 and Stx2 are highly related in both structure and function but cannot be neutralized by heterologous antisera. Although STEC that produce either Stx1 or Stx2 may cause severe disease in people, strains that make Stx2 are more likely to cause HUS (21). However, it is not clear how Stx gains access to the bloodstream from the intestinal lumen. Nevertheless, the preponderance of evidence indicates that Stx targets small vessel endothelial cells, and that toxin tropism, either directly or indirectly, leads to HUS (22).

Cell-surface expression of the Stx receptor, globotriaosylceramide (Gb3), is required for cytotoxicity in vitro (23), and, in some models of STEC infection, Stx damages the intestinal epithelia (18, 19). Historically, it was presumed that Gb3 is not present on the gut epithelia (24, 25), and, thus, the mechanism by which the toxin might gain systemic access was unclear. However, contrary to this perception, we recently demonstrated that small amounts of Gb3 synthase mRNA and Gb3 are in fact found in human intestinal tissue (26).

In this study, we hypothesized that Stx uptake from the gut, as well as the degree of kidney damage (and thus susceptibility to HUS) following infection might, in part, be mediated by diet. Specifically, we theorized that a diet or an intestinal commensal population that promoted the production of butyric acid in the gut would increase susceptibility to STEC infection. Butyric acid is normally a beneficial catabolite of intestinal fiber fermentation that promotes cellular health (27) and acts as the primary energy source for colonic enterocytes (28, 29). Nevertheless, when cultured cells are treated with butyric acid, the surface expression of Gb3 increases, which, in turn, leads to enhanced sensitivity of those cells to the cytotoxin Stx (30–33). To test our diet hypothesis, we fed mice a high fiber diet (HFD), which in turn, increased intestinal butyrate levels, and then asked whether the response of those animals to E. coli O157:H7 infection differed from mice fed a low fiber diet (LFD). We found that an HFD increased Stx binding to colonocytes and renal tissues of mice, reduced the level of commensal E. coli, and increased the susceptibility of mice to E. coli O157:H7 infection and disease.

Results

Influence of Butyric Acid on Stx Receptor on Human Colonic Epithelial Cells.

To confirm that butyrate enhances Gb3 levels on colonic epithelial cells, we grew human colonic epithelial (HCT-8) cells in media with or without sodium butyrate and assayed for cell-surface Gb3 by flow cytometry. Gb3 levels on butyrate-exposed HCT-8 cells increased 10-fold (Fig. 1A). In addition, HCT-8 cells maintained in butyrate exhibited 1,000-fold higher sensitivity to Stx1 intoxication compared with control cells (Fig. 1B). This enhancement of Gb3 levels on HCT-8 cells occurred within 24 h of butyrate treatment; however, within 4 h of butyrate removal, cell-surface Gb3 levels returned to basal levels (Fig. 1C). The rapid responsiveness of cell-surface Gb3 to butyrate suggests a dynamic temporal fluctuation in Gb3 that varies with local butyrate concentration.

Fig. 1.

Butyric acid-enhanced Gb3 expression on human colonic epithelial cells. (A) Flow cytometric analysis of HCT-8 cells treated with 5 µM sodium butyrate for three passages and incubated with a monoclonal antibody against Gb3. The horizontal line indicates the gating window used to quantify Gb3-positive cells. (B) Viability of PBS- or butyrate-treated HCT-8 cells after exposure to Stx1. (C) Cell-surface Gb3 (green) detected on HCT-8 cells 2 or 4 h after cells were washed and butyrate thus removed. (Magnification: 400×.) Evan’s Blue (0.01%) was used as a counterstain (red). Arrow identifies Gb3 expression in untreated HCT-8 cells. Control cells were treated with PBS only.

Diet-Mediated Increase in Butyric Acid Influences Gb3 Expression in Vivo.

We next asked whether increased butyrate levels would enhance Gb3 expression in the mouse gut. To modulate the intestinal butyrate level in mice, we fed either an HFD or an LFD to BALB/c mice. We used guar gum as the fiber source in the two diets (Table S1) because it is highly fermentable in the gut (34). Mice that ate either diet appeared equally healthy (Fig. S1). We collected stool and cecal contents from HFD- or LFD-fed mice and quantitated the butyrate levels in the samples by gas chromatography and mass spectrometry. Stool and cecal contents from HFD-fed mice consistently contained more butyrate than similar samples from LFD-fed mice (Fig. 2A). These data are consistent with findings from human subjects (35, 36). Additionally, because stool from mice fed either diet contained less butyrate than the cecal contents (Fig. 2A), we hypothesized that some butyrate was absorbed systemically, a possibility supported by studies done in pigs (37). Furthermore, we detected more Gb3 on intestinal tissue from HFD-fed mice than on similar samples from mice on a typical 4% fiber diet (Fig. 2B). Because butyric acid enhances Gb3 expression on human proximal tubule (30) and glomerular epithelial cells (31) in vitro, and it was likely that butyrate was absorbed systemically in HFD-fed mice, we asked whether those mice had greater kidney Gb3 staining than did LFD-fed animals. Although sections of kidney from HFD-mice exhibited similar patterns of Gb3 staining compared with those from LFD-mice (Fig. 2C), there were numerous areas of increased intensity of Stx binding in tissues of HFD-fed mice (Fig. 2C, arrows). The increase in Gb3 expression on kidney tissue of mice fed an HFD was statistically significant, as determined by fluorescence intensity quantitation (Fig. 2D).

Fig. 2.

An HFD increased intestinal butyrate and Gb3 expression in mice. (A) Quantity of butyrate in stool and cecal contents of HFD- or LFD-fed mice. (B) Gb3 expression on intestinal tissue from mice fed an HFD (detected with anti-GB3 antibody) at 100× magnification and (C) on kidney sections (detected with an Stx1 probe). Arrows indicate regions of increased signal intensity. (Magnification: 100×.) (D) Quantitation of fluorescence intensity from C by ImageJ software (65). Error bars represent the standard error of the mean (SEM).

Diet Alters Susceptibility of Mice to Infection with E. coli O157:H7.

To determine the effect of an HFD on susceptibility to STEC infection, we fed two groups of 10 mice each either an HFD or LFD for 2 wk and then infected them with 10¹¹ cfu E. coli O157:H7. Some of the infected mice in the HFD group consistently showed signs of greater morbidity, such as lethargy, ruffled fur, and weight loss by day 7 (Fig. 3A). In most cases, weight loss that was greater than 20% of body weight led to death (Fig. 3B). Over the course of four independent infection experiments, there were 12/36 (33%) deaths in the HFD group and 4/34 (12%) in the LFD mice. Additionally, mice that consumed an HFD maintained greater levels of E. coli O157:H7 than did the LFD-fed animals (Fig. 3C). We examined sections of the intestines and kidneys from infected HFD-fed mice that lost >20% body weight (Fig. S2) for the presence of lesions and compared those to samples from LFD-fed mice that had been infected for the same amount of time (Fig. 3D, Fig. S2). Neither group of mice had intestinal lesions; however, kidneys from the HFD-fed mice displayed moderate acute tubular necrosis (Fig. 3D, multiple arrows) whereas those fed the LFD had only mild lesions (see single arrow).

Fig. 3.

Mice fed an HFD are more susceptible to infection with E. coli O157:H7. (A) Average weight loss in HFD-fed (triangles) and LFD-fed (squares) mice infected with STEC (n = 10). Error bars indicate SEM. *P < 0.001, **P < 0.01, ***P < 0.05. (B) Mortality of STEC-infected HFD-fed (triangles) or LFD-fed (squares) mice (n = 10). (C) Geometric mean of colonization by O157:H7 in HFD-fed (triangles) and LFD-fed (squares) mice (n = 10). Error bars indicate 95% confidence interval. *P < 0.001. (D) Intestinal and kidney tissue sections stained with hematoxylin-eosin (H&E) from infected mice. (Magnification:100×.) Arrows indicate areas of acute tubular necrosis.

To determine whether the greater susceptibility of the HFD mice to E. coli O157:H7 was specific to the presence of Stx2, we infected mice on an HFD or LFD with either the wild-type 86-24 or an isogenic toxin minus mutant and monitored the animals for clinical signs of disease. We found that whereas 7/8 HFD and 3/8 LFD mice infected with 86-24 died, no mice lost weight (Fig. S3) or succumbed to infection with the Stx2⁻ 86-24 regardless of the diet they were fed.

Diet Influences the Composition of the Intestinal Microbiota.

Multiple factors influence the microbiota of the intestine and thus the luminal milieu of metabolites; these variables include age (38), disease state (39), and diet (38, 40). To investigate the potential role of the intestinal microbiota on the susceptibility of the HFD-fed mice to E. coli O157:H7, we determined the gut microbiota composition in mice fed a typical 4% fiber diet (day 0) and then switched to either an HFD (10% fiber) or LFD (2% fiber) for 14 d. We found that mice on an HFD had reduced numbers of commensal Escherichia_Shigella (hereafter Escherichia) spp compared with LFD-fed mice on day 14 post diet change (Fig. 4A, Fig. S4), a finding consistent with studies in humans (35, 36) and piglets (41). This reduction in Escherichia spp suggests that an HFD may create an available niche for the O157:H7 to colonize, despite the overall similarity between the gut communities in the LFD and HFD groups at day 14 (Fig. S5). Despite the reduction in Escherichia spp in HFD-fed mice, overall numbers of intestinal commensals appeared to be increased compared with LFD animals as shown by an increased quantity of bacterial DNA in the stool (Fig. 4B) and by more bacteria in the lumen of the colon of H&E-stained tissue sections from most animals (Fig. 4C). Thus, an HFD may create an intestinal environment that provides an advantage for STEC to colonize and cause disease.

Fig. 4.

An HFD alters the intestinal commensal composition. (A) Relative number of Escherichia operational taxonomic units (OTU) in stool of mice on an HFD or LFD as determined by 16S-based profiling of bacteria present in fecal pellets. (B) Total DNA (isolated to enrich specifically for bacteria DNA) present in fecal pellets of mice on different diets. (C) H&E stain (40×) of a cross-section of intestine from mice fed an LFD or an HFD. An Inset in the HFD-fed intestinal section is shown adjacent to that section at 1,000× with the luminal bacteria in blue.

Discussion

Here, we demonstrated in vivo a phenomenon that was previously shown only in tissue culture. Specifically, we found that butyric acid, generated in the gut because of an HFD, increased the expression of the Stx receptor, Gb3, on the colonic epithelium and in renal tissues. We further established that the butyric acid-mediated increase in Gb3 levels was temporally dynamic. The amount of Gb3 was reduced to prestimulation levels within hours of the withdrawal of butyric acid. Finally, we showed in a mouse model that not only did the HFD result in increased intestinal butyrate levels, but it also led to a reduction in the resident Escherichia spp. The HFD-fed mice exhibited greater colonization, morbidity, and mortality after STEC infection than did mice on an LFD. Furthermore, the greater susceptibility to the HFD was related specifically to the presence of Stx2. In all, we demonstrate here that an HFD can modulate Stx receptor levels, which, in turn, can exacerbate the outcome of STEC infection.

Our finding that Gb3 levels were temporally dynamic in response to butyrate suggests that Gb3 expression on tissues may fluctuate with diet and within microenvironments of the tissue as a function of local butyrate exposure. This latter hypothesis may explain the lack of Gb3 detection on colonic tissues in previous studies (24, 25, 42) in which butyrate may have been incidentally removed during bowel preparation for surgery. Mechanical bowel preparation with one or more types of laxative up to 24 h before surgery is the standard procedure for patients who undergo elective colorectal procedures (43). In addition to the enhanced expression of Gb3 on intestinal tissues of HFD-fed mice, we also noted increased Gb3-staining intensity in renal tissues from those same animals. The higher levels of Gb3 in the kidneys of HFD-fed mice was presumably a consequence of increased systemic levels of butyrate taken up from the gut. Because the pattern of renal Gb3 expression in mice fed an HFD was similar to that in the kidneys of LFD-fed mice, we propose that the butyrogenic effect on Gb3 expression occurs specifically in those cells that have the capacity to express Gb3.

That butyric acid and an HFD have a negative effect on the host with respect to STEC infection is contrary to their essential role in normal colonic health (44, 45). Butyrate accounts for up to 70% of the energy source for colonic enterocytes (46, 47). It also is important in maintaining normal physiology and morphology of the epithelia in the gut (28) and has been proposed for use as a probiotic in cases of chronic colitis (48). Furthermore, there is a link between reduced gut butyrate metabolism and colitis in mice (49) and humans (28, 44). Finally, at least two studies show beneficial effects of butyrate during infection with Salmonella (50, 51). In contrast, this study suggests that butyrate can be detrimental to the host under certain conditions.

From our results, we cannot rule out that butyric acid sensitized tissues to Stx by means other than enhancement of Gb3 expression. It is clear that butyric acid has a profound effect on cellular physiology (29) and on the cellular transcriptional program (27). Butyric acid also affects cellular sensitivity to another ribotoxin, ricin (52), which does not use Gb3 as a receptor (53). Additionally, Lingwood et al. showed that butyric acid may alter the trafficking of Stx in cells to promote the interaction of toxin with its cellular target, the endoplasmic reticulum (54). Furthermore, it is possible that butyrate has a direct effect on the bacterium itself. We did not explore this possibility although one group showed that including butyrate in the growth medium of an STEC culture resulted in increased adherence and microcolony formation on CaCo₂ cells (55).

It is also possible that other unidentified factors of an HFD, in addition to butyrate production, contribute to enhanced susceptibility to STEC infection. However, in stool samples analyzed from mice on an LFD or an HFD, the other major small chain fatty acid metabolites produced in the gut, propionic acid and acetic acid, were at similarly low concentrations.

We were surprised to find a mostly analogous microbiota composition in the guts of HFD- and LFD-fed mice (Fig. S5). A significant difference in the two groups, however, was the finding of higher numbers of commensal Escherichia spp in the LFD-fed group, despite an overall reduction in gut commensal bacteria in this group. A similar decrease in Escherichia spp has been observed in humans (36) and piglets (41) that consumed an HFD. However, we consistently observed higher colonization by E. coli O157:H7 in the HFD-fed mice than the LFD-fed mice throughout the course of infection. We postulate that STEC colonize to a higher degree in HFD- than LFD-fed mice in three ways. First, because of lower levels of commensal Escherichia spp in the HFD gut, there may be reduced competition against pathogenic Escherichia spp. Second, the overall lower quantity of commensal bacteria in the LFD mice suggests that an HFD is more favorable for colonization in general. Third, given that Stx enhances intestinal colonization during STEC infection (56), the increase in intestinal Gb3 expression due to the HFD may promote colonization as well.

Taken together, our results suggest that an HFD may pose a two-pronged risk for those exposed to STEC such as E. coli O157:H7. First, butyrate produced under high fiber conditions leads to enhanced Gb3 levels on the intestinal cell surface and, after absorption into the blood stream from the gut, on kidney cells. Such higher colonic and renal Gb3 levels may enhance systemic transport of Stx from the gut and increase Stx sensitivity of kidney cells. Second, an HFD causes a reduction in native Escherichia spp, which, in turn may promote enhanced colonization by incoming pathogenic STEC. Our model is consistent with a study in cattle in which a higher fiber grain diet in cattle led to greater colonization by STEC (57).

In summary, our model addresses two unresolved facets of STEC pathogenesis: (i) the mechanism of Stx uptake in the gut; and (ii) the disparity in HUS cases following STEC infection. It is clear that intestinal ecology and diet likely result in microenvironments with variable levels of butyrate and Gb3. Therefore, individual hosts likely have different Gb3 levels on and within their gastrointestinal tracts, and those variations in receptor quantities could result in disparities in the amount of Stx delivered systemically. Higher levels of Stx in the host presumably lead to an increased risk of HUS. Besides diet, host age is a factor that influences microbiota composition (39). In addition, in rabbits there is an age-related change in the expression of Gb3 in tissues (58). Perhaps an age-dependent change in the host microbiota alters the butyrate-producing potential in the gut, which then influences Gb3 levels. Such differences in microbiota, and thus alterations in butyrate and Gb3, could explain some of the stark differences in HUS susceptibility following infection in children <10 y old compared with adults.

Materials and Methods

Cell Culture.

HCT-8 cells were obtained from ATCC and grown in RPMI-1640 media (ATCC) supplemented with 10% (vol/vol) heat-inactivated fetal bovine serum (FBS, Life Technologies). Vero cells (ATCC) were maintained in eagle's minimal essential medium (EMEM, Cell Gro) supplemented with 10% (vol/vol) heat-inactivated FBS and supplemented with 10 mM l-glutamine (Lonza), 10,000 units/10,000 µg/mL penicillin/streptomycin (Life Technologies), and 50 mg/mL gentamicin (ATCC). Cells were passaged at ∼80% confluency every 3–4 d.

Antibodies.

To detect Gb3 in tissue sections and on cells in culture, we used a rat monoclonal IgM to human CD77 (Gb3), clone [38-13], (Genetex). The cognate secondary antibody used to detect bound Gb3 antibody was goat anti-rat IgM Alexa Fluor 488 (Life Technologies). To detect Stx1 on tissues and on cells, we used a rabbit polyclonal antibody developed in our laboratory (59). The cognate secondary antibody used to detect antibody against Stx1 or Stx2 was goat anti-rabbit IgG Alexa Fluor 488 (Life Technologies). Sodium butyrate was purchased from Sigma-Aldrich. Guar gum used in custom fiber diets was purchased from Sigma-Aldrich.

Flow Cytometric Analysis.

HCT-8 cells maintained in RPMI-1640 (ATCC) and 5 µM butyrate (Sigma-Aldrich) and grown in six-well dishes to ∼80% confluency were washed with PBS and then detached from the plate with 0.5 mL of 0.05% trypsin/EDTA (Cell Gro) for 5 min at 37 °C. Cell clumps were then removed gently with a 25G needle/syringe followed by transfer through a 70-µm nylon filter. Cells were transferred to a 96-well v-bottom polypropylene plate (Becton Dickinson), and the plate was subjected to centrifugation at 400 × g at 4 °C for 4 min. Primary anti-Gb3 rat monoclonal antibody (1:50 in 3% BSA/PBS; Matreya) was added to the cells, and the plate was incubated at 4 °C for 1 h. The cells were then washed with PBS, and secondary goat-anti-rat AF488 (1:1,000 in 3% BSA/PBS; Life Technologies) was added. The plate was then incubated at 4 °C in the dark. Following a final PBS wash of the cells, they were resuspended in 400 µL of PBS and immediately analyzed with an LSR II Flow Cytometer (Becton Dickinson).

Purification of Stx.

Stx1 was purified as previously described (60) by a single-step immunoaffinity procedure.

Stx Cytotoxicity Assay.

A cytotoxicity assay modified from Gentry and Dalrymple (61) was carried out on HCT-8 cells in the presence or absence of sodium butyrate. HCT-8 cells were seeded at 1 × 10⁴ per well of a 96-well plate (Corning) and grown for 48 h at 37 °C in 5% CO₂. The cells were then exposed to 10-fold serial dilutions of a stock of Stx1 at 25 ng/mL, prepared in RPMI-1640 tissue culture medium that contained 10% FBS for an additional 48 h of incubation. Control cells were given medium alone. Cells were then fixed in Formaldefresh (Fisher Scientific) and stained with 1.3% crystal violet (Sigma-Aldrich) in 5% ethanol (Fisher Scientific). The optical density (OD) at 630 nm of fixed, stained cells in each well on a plate was determined in a microplate reader. The average OD₆₃₀ of three replicates for each sample was calculated. The butyrate-treated group and the control group were each independently normalized to the OD₆₃₀ value of the Stx1-negative samples.

Animal Studies.

All animal work was approved by the Institutional Animal Care and Use Committee at the Uniformed Services University. An intact-commensal flora model (62) in female BALB/c mice (Charles River) was adapted for these studies. Mice were weighed and then introduced to one of two custom diets: an HFD (10% guar gum) or an LFD (2% guar gum) (Harlan Laboratories). Control animals were maintained on the standard diet (Teklad Global Protein 18% with 4% soluble fiber; Harlan Laboratories). The specifications for each custom diet are listed in Table S1. Mice were acclimated to each diet for 2 wk and then infected by gavage with 10⁹ to 10¹¹ cfu of E. coli strain O157:H7, isolate 86-24 (Stx2+) that was originally obtained from Phil Tarr (Washington University, St. Louis, MO). Alternatively, some mice were infected with an isogenic stx₂ mutant of 86-24 (TUV86-2) provided to us by A. Donahue-Rolfe (63). To prepare the inoculum, the bacteria were cultured in LB media overnight at 37 °C with shaking at 225 rpm and then pelleted by centrifugation for 10 min at 1,000 × g. The organisms were then resuspended in 1/40th volume of PBS. One hundred microliters of culture was gently gavaged into each mouse. The day of infection was designated as day 0 in each study. After mice were infected, they were weighed daily for 2 wk. They were also monitored for other signs of morbidity such as ruffled fur and lethargy. To determine the level of colonization in the gut of infected mice, the amount of E. coli O157:H7 in fecal pellets was assessed as follows. Fecal pellets from individual mice were collected, weighed, and suspended 1:10 (wt/vol) in PBS. The fecal pellet/PBS mixtures were mixed by shaking on a vortex at room temperature for 1 min, every 10 min, for 30 min. Ten-fold dilutions of the mixtures were made in PBS, and 100-µL aliquots of the homogenates were plated onto Sorbitol–MacConkey agar plates. Plates were incubated overnight at 37 °C. Colony counts were determined the next day.

Harvesting and Sectioning of Mouse Intestines and Kidneys.

Intestinal sections were immediately removed from mice following death and either submerged in formalin to fix the tissue or embedded and frozen in Tissue-Tek Optimum Cutting Temperature (OCT) compound (Sakura Finetek) unfixed. For histopathological evaluation of tissue sections, hematoxylin-stained fixed sections were deemed optimal for analysis. To minimize background staining for detection of Gb3 on tissues, unfixed frozen tissues were found to be superior to fixed tissue.

Immunofluorescence of Stained Tissues and Cells.

Immunostaining of intestinal tissue sections or HCT-8 cells for Gb3 was done as follows. An ∼2-cm length of intestine distal to the cecum was removed from a mouse that had been killed. The luminal contents of that tissue piece were gently flushed out with PBS contained in a syringe fitted with a pipet tip. The intestinal segment was cut longitudinally with scissors to expose the lumen. Tissue was positioned on semisolid OCT medium (Sakura Finetek) in a sectioning boat and covered in liquid OCT, then frozen on dry ice. From the tissue block, 6-μm sections of tissue were sliced onto positively charged Superfrost Plus slides (Fisher Scientific), and slides were kept frozen on dry ice. Slides that were to be stained were thawed and fixed for 5–10 min in cold acetone (−20 °C). Excess acetone was removed, and the specimen was encircled in a hydrophobic barrier with a Papanicolaou (PAP) pen (Polysciences). To prepare HCT-8 cells for staining, cells were grown in wells of eight-chamber slides. The slides were then washed with PBS and fixed in acetone (−20 °C) for 10 min. The slides were then allowed to air dry. Slides with fixed intestinal sections or HCT-8 cells were then incubated for 30 min with 3% BSA in PBS as a blocking reagent, washed in PBS, incubated with an anti-Gb3 rat monoclonal antibody (Matreya), diluted 1:50 in 3% BSA/PBS for 1 h, and then washed again with PBS. Goat anti-rat AF488 (Life Technologies) secondary antibody diluted in 3% BSA/PBS to 1:1,000 was then added to the slides for 1 h, and then slides were washed in PBS. To counterstain HCT-8 cells on the slides, 0.01% Evan’s Blue dye in PBS was added to the slides for 30 s followed by a final wash with PBS. Intestinal tissue sections on slides were mounted with anti-fade glycerol solution (Life Technologies) and observed with an Olympus BX60 fluorescence microscope.

For immunostaining of mouse kidney sections, the presence of Gb3 was detected indirectly with Stx1 as a probe of unfixed frozen tissue sections by a previously described method (64). Briefly, kidneys were removed from mice and immediately frozen in liquid OCT. Sections (6 µm) prepared on positively charged slides were air dried overnight at room temperature to fix the tissue. Tissue sections were then blocked for 30 min in 1% normal goat serum (NGS) in PBS. Excess NGS was removed, and 200 μL of Stx1 (200 ng/mL) was added, followed by incubation of the slide at room temperature for 1 h. The slide was then washed with PBS, and 200 μL of rabbit polyclonal anti-Stx1 (59) diluted to 1:50 in 1% NGS/PBS was added for 1 h. The kidney section on the slide was washed again with PBS, and secondary goat anti-rabbit AF488 diluted to 1:500 in 1% NGS/PBS was added to the slide for 1 h. The tissue section was then subjected to a final PBS wash and mounted and observed as described above.

Quantitation of Immunofluorescence.

Image J software (65) was used to quantitate fluorescence intensity. Ten random positively stained kidney tubules (Fig. 2C) from each tissue were quantitated by calculating the ratio of fluorescence intensity to the area detected.

Butyrate Extraction and Gas Chromatography/Mass Spectroscopy.

The butyrate extraction procedure was adapted from the method of Richardson et al. (66). Fecal pellets or cecal contents were collected from mice, weighed, and suspended in 2 mL of PBS in glass vials (Fisher Scientific) with screw caps. The vials were immediately stored at −20 °C. To extract butyrate from the samples, the vials were thawed at room temperature, and 5 mM ethyl-butyric acid (Sigma-Aldrich) in dimethyl sulfoxide (Sigma-Aldrich) was added to each vial as an internal reference. The vial contents were then mixed by shaking on a vortex for at least 2 min until the samples within them were homogenous. The largest solid particles were then allowed to settle in the vial, and a 1-mL homogenate was then transferred from each vial to a 5-mL glass screw-cap tube (Fisher Scientific). To acidify the samples, 50 µL of 5 M HCl (Sigma-Aldrich) was added to each tube. Next, 2 mL of diethyl ether (Sigma-Aldrich) was added to each tube. The tubes were then shaken on a vortex for 2 min, and then they were subjected to centrifugation for 10 min at 3,000 × g. Aliquots of each resultant supernatant (1 µL) were separated on a Nukol Supelco capillary GC column; 30 m × 0.25 mm × 0.25 µm (Sigma-Aldrich). Helium was used as the carrier gas at 0.80 mL/min from 100 °C–200 °C at 8 °C/min. A mixture of standard free fatty acids (Sigma-Aldrich) was used to generate a standard curve and to identify the appropriate peak by mass spectroscopy that corresponded to butyric acid.

Immunohistochemistry.

Tissues were removed, washed in PBS, and immediately submerged in 10% neutral buffered formalin, pH 6.8–7.2 at 25 °C (Fisher Scientific). Tissues were embedded in paraffin and sectioned at 6 μm onto glass slides. For hematoxylin and eosin (H&E) staining, sections were deparaffinized in ethanol and stained by standard protocol. Tissues were sectioned and stained by the Biomedical Instrumentation Center, Uniformed Services University of the Health Sciences, Bethesda, MD and by Histoserv, Inc., Gaithersburg, MD.

Histopathological Analysis of Tissue Sections.

To assess tissues for the presence of lesions, a veterinary pathologist evaluated sections in a blinded manner.

Mouse Gut Microbiota Sequencing Experiment.

The mice used for the gut microbiota sequencing experiment were housed individually to prevent mouse-to-mouse transmission of endogenous microbiota. Ten mice fed either an HFD or an LFD were included in this study. Fecal pellets were collected from mice in cages that contained a wire grid to limit the contamination of fecal samples with microbiota present on skin and fur. Fecal pellets were collected before starting the HFD or LFD, and then 7 and 14 d after the start of each diet. On day 15, mice were infected by gavage with 1 × 10¹¹ cfu of E. coli O157:H7. Fecal pellets from each mouse were weighed, and bacterial DNA was isolated from the pellets by the QiaAmp DNA Stool extraction system (Qiagen).

16S V6 Hypervariable Region PCR, Illumina Sequencing, and Subsequent Microbial Community Profiling.

The V6 region of the bacterial 16S rRNA gene was amplified by the PCR from fecal pellet DNA with a combination of five forward (CNACGCGAAGAACCTTANC, CAACGCGAAAAACCTTACC, CAACGCGCAGAACCTTACC, ATACGCGARGAACCTTACC, and CTAACCGANGAACCTYACC) and four reverse primers (CGACAGCCATGCANCACCT, CGACAACCATGCANCACCT, CGACGGCCATGCANCACCT, and CGACGACCATGCANCACCT) (67–69). V6 PCRs were performed in triplicate for each sample, and the replicate reactions were pooled before sequencing. V6 amplicons were sequenced on an illumina Hi-sEq (100-bp single-end run) at the Tufts University Core Facility. Raw reads were processed in Galaxy (70) as follows: illumina adapters, primer sequences, and low quality bases at the ends of each read were removed; any sequences with an “N” base were removed from the dataset; sequences were filtered to remove any reads with a quality score of less than 15 for 95% of the read. All subsequent sequence analyses were done with mothur (71). To simplify the dataset and reduce computational time, the sequences from each sample that remained after initial processing were randomly subsampled to a depth of 115,000 reads per sample. To increase efficiency and accuracy, sequences were aligned to a V6-specific, curated database (72) derived from the full-length SILVA alignment. The alignment was screened to remove sequences that were shorter than 56 bp or greater than 72 bp in length as well as any sequences that contained homopolymer runs of >4 bp. The resulting alignment was then filtered to remove any columns that contained missing information. Reads were preclustered so that any sequences that contained a single base pair difference were considered to be the same. Preclustered reads were classified with the mothur implementation of the RDP Bayesian classifier (73) using a cutoff of 60% bootstrap support over 100 iterations. Sequences from mitochondria, cyanobacteria, chloroplasts, and sequences that were classified as “unknown” or “unclassified” at the phylum level were removed from the dataset. The remaining aligned sequences were used to generate a pairwise distance matrix and clustered into operational taxonomic units (OTUs) using average-linkage clustering at an identity of 97%. OTUs were classified with the mothur implementation of the RDP classifier (73) as described above. To compare the intrasample diversity of each of the murine gut communities, we normalized the sequences from each sample based on the lowest number of sequences found in any of the samples. This workflow provided adequate sample coverage (>0.97) as determined by Good’s estimate (74). Rarefaction curves were generated by plotting the mean number of OTUs from all animals in each experimental group (Fig. S6). Coordinates for nonmetric multidimensional scaling plots were generated using mothur and plotted in Microsoft Excel. A complete list of OTUs identified in this study is given in Dataset S1.

Statistical Analyses.

Prism v4 (Graphpad Software) was used to analyze data throughout the manuscript. SEM was calculated for replicate samples in Figs. 2 A and D, 3A, and 4B, and Fig. S3. A Student t test was used to determine statistical significance between groups in Figs. 2 A and D and 4B. A Two-way ANOVA analysis with two-way repeated measures and the Bonferroni posttest correction was performed on data shown in Fig. 3A and Fig. S3 to determine statistical significance between time-matched samples. To determine statistical significance in Fig. 3C, a two-way ANOVA analysis with two-way repeated measures and the Bonferroni posttest correction was performed on the log-transformed individual values. Because the geometric mean was plotted in Fig. 3C, the error represented among replicate samples indicates the 95% confidence interval of time-matched values between each group. Analysis of molecular variance (AMOVA) (75, 76) was used to compare the intestinal bacterial community structures of day 0, day 14 LFD, and day 14 HFD mice, shown in Fig. S5.