Prolonged Impact of Antibiotics on Intestinal Microbial Ecology and Susceptibility to Enteric Salmonella Infection^▿

Bacterial strains and growth conditions.

Salmonella enterica serotype Typhimurium strain ATCC 14028 cells (ATCC) were cultured aerobically at 37°C in Luria-Bertani (LB) broth as described previously (34). Briefly, Salmonella bacteria were grown in a 10-ml standing culture for 3.5 h at 37°C to select for the most motile bacteria. The top 5 ml of the culture was transferred into 45 ml of prewarmed LB broth and grown with shaking for 1.5 h at 37°C. The quantity of Salmonella cells was determined using a Petroff-Hauser chamber (Hauser Scientific, PA), and the appropriate dose was prepared in sterile 0.2 M phosphate buffer (pH 8).

Animals.

Female FvB mice were obtained from Taconic Laboratories at 4 weeks of age. Animals were housed under specific-pathogen-free conditions in the Medical College of Wisconsin Biomedical Resource Center vivarium. The husbandry and diet for all study animals were controlled and unchanged through the course of this study to reduce the variable environmental impact on the intestinal microbiome. The animal care and use committee at the Medical College of Wisconsin approved all animal-related experiments and procedures.

Antibiotic treatment experiments.

The following antibiotics were used: streptomycin sulfate (Fisher), bacitracin (Fluka), vancomycin hydrochloride hydrate (Sigma-Aldrich), ampicillin trihydrate (Sigma-Aldrich), neomycin sulfate (Fisher), and metronidazole (Sigma-Aldrich). Four different antibiotic regimens were used to treat mice: control (no antibiotics), streptomycin (0.5 g/250 ml drinking water), streptomycin-bacitracin (0.5 g of each/250 ml drinking water), and vancomycin (0.125 g)-neomycin (0.25 g)-metronidazole (0.25 g)-ampicillin (0.25 g, combined in 250 ml water). These are drugs and combinations that have been commonly used to disrupt intestinal microbial ecology (5, 8, 31, 47). Streptomycin, used most commonly in animal infection models, is not effectively absorbed from the GI tract; streptomycin is most effective against gram-negative bacteria but also has some activity against gram-positive bacteria and Mycobacterium tuberculosis. Bacitracin is also poorly absorbed from the gut and is effective primarily against gram-positive bacteria. The four-drug regimen of ampicillin, vancomycin, neomycin, and metronidazole (AVNM) demonstrates broad-spectrum coverage. Ampicillin is moderately well absorbed from the gut but is rapidly eliminated systemically and shows activity against gram-positive and -negative aerobes and anaerobes. Metronidazole is well absorbed from the gut but is rapidly cleared from systemic circulation and shows activity against gram-positive and -negative anaerobes. Neither vancomycin nor neomycin is well absorbed by the GI tract. Vancomycin has activity against gram-positive bacteria, while neomycin has broad-spectrum activity but is particularly effective against gram-negative bacteria. At 5 weeks of age, groups of mice (five mice per group) were treated with antibiotics, as specified, in their drinking water for 7 days. Fecal pellets were obtained from the mice for aerobic and anaerobic culture. Animals were sacrificed, and intestinal tissue was taken for histology, aerobic and anaerobic culture, fluorescence in situ hybridization (FISH), total bacterial genomic isolation, and quantitative PCR (qPCR).

Antibiotic recovery experiments.

Groups of mice (five mice per group) were treated with streptomycin-bacitracin in their drinking water for 7 days as described above and then placed on regular water without antibiotics. Age-matched groups of control mice received no antibiotics in their drinking water. Groups of control animals and treated animals were sacrificed at 1, 3, 5, 7, 14, and 21 days after the cessation of antibiotic treatment. The intestinal tracts were removed, divided, and analyzed as described above. The experiment was repeated, and results were reproducible.

Bacterial culture of feces and large-intestinal contents.

Fresh stool pellets or large-intestinal tissue samples were collected from mice from each treatment group. The specimens were then homogenized in 2.0 ml of 1× phosphate-buffered saline (PBS). The homogenized samples were then plated onto LB agar for aerobic bacterial growth and onto Schaedler agar for anaerobic bacterial growth as previously described (16, 28). The LB plates were incubated at 35°C for 24 h, and photographs of the resulting growth were taken. The Schaedler plates were preincubated in a GasPak anaerobic chamber (BD Diagnostic Systems) for 18 h, and after plating, they were placed into a GasPak anaerobic chamber and incubated at 35°C for 48 h. Photographs were taken after incubation.

Salmonella infection experiments.

As in the antibiotic treatment experiments described above, groups of mice (five mice per group) were treated with antibiotics, as specified, for 7 days in their drinking water. Control mice received water without antibiotic supplementation. On day 7, the antibiotics were withdrawn from the treatment group, and all mice received untreated water. Mice were deprived of food overnight. Each group of animals was inoculated with 10⁸ CFU of Salmonella by intragastric gavage. Mice were sacrificed 3 days postinoculation to allow a measurable translocation of Salmonella to the liver and spleen but prior to the animals becoming moribund. Intestinal tissue was isolated and processed as described above for the antibiotic experiment. Spleen and liver were removed and analyzed for the presence and abundance of translocated Salmonella cells. Salmonella burden was determined by homogenization of spleen and liver in sterile PBS and plating in dilution onto Salmonella-Shigella agar. The experiment was repeated, and results were reproducible.

Salmonella infection after antibiotic recovery.

Groups of FvB mice (five mice per group) were given untreated drinking water or drinking water containing streptomycin-bacitracin for 1 week as described above. Antibiotic-treated animals and control animals were inoculated with 10⁸ Salmonella enterica serovar Typhimurium cells by oral gavage 1, 3, 5, 7, 14, and 21 days after antibiotic withdrawal, as described above. Three days postinfection, animals were sacrificed and analyzed as described above. The experiment was repeated, and results were reproducible.

FISH.

FISH was performed on mouse terminal ileum and cecum as described previously (4). Briefly, intestinal tissue from each mouse was fixed in Carnoy's fixative, 3-μm sections were mounted onto slides, and FISH was performed using a combination of a 6-carboxyfluorescein (FAM)-labeled oligonucleotide probe for segmented filamentous bacteria (SFB) (SFB1008 [FAM-GCGAGCTTCCCTCATTACAAGG]) (37) and a Texas Red-labeled universal bacterial probe (Bact338 [Texas Red-GCTGCCTCCGTAGGAGT]) (4) (Operon Technologies, Huntsville, AL). Slides were viewed by fluorescence microscopy using a Nikon E400 upright microscope. Images were captured using a Photometrics CoolSnap ES charge-coupled-device camera and analyzed using Metaview software (Universal Imaging Corporation, Molecular Devices). Representative sections from mice belonging to each study group were photographed.

Histology.

Three-micrometer sections of zinc formalin- or Carnoy's fixative-fixed distal small intestine (DSI), cecum, and large intestine (LI) were mounted onto slides and stained with hematoxylin and eosin. Slides were examined by an anatomic pathologist (N.H.S.) using a Nikon E400 upright microscope. Images were captured using a Spot camera and analyzed using Spot software, version 3.5.4 (Diagnostic Instruments, Inc.). Cecal pathology was scored in a blinded fashion, grading the extent of edema (0, no edema; 1, less than 50% of mucosa involved; 2, more than 50% of mucosa involved; 3, total mucosal involvement), inflammation (0, no acute inflammation; 1, focal acute inflammation in lamina propria; 2, extensive submucosal neutrophilic infiltrate; 3, transmural neutrophilic infiltrate), and hyperplasia (0, no epithelial hyperplasia; 1, twofold increase in thickness; 2, threefold increase in thickness; 3, fourfold or greater increase in thickness). Cecal sections from each mouse (five mice per group) were scored for each criterion and combined for a total enteritis score.

Bacterial genomic DNA extraction.

The DSI, cecum, and LI isolated from the experimental animals were weighed and then homogenized using a Polytron PT 10-35 homogenizer (Kinematica Switzerland) in 2 ml sterile PBS. Bacterial genomic DNA was extracted from the DSI, cecum, and LI using the Qiagen stool kit according to the kit directions, using the optional high-temperature step.

Quantitative real-time PCR amplification of 16S rRNA gene sequences.

The abundances of specific intestinal bacterial groups were measured by qPCR using the MyiQ single-color real-time PCR detection system (Bio-Rad, Hercules, CA) using group-specific 16S rRNA gene primers (Operon Technologies, Huntsville, AL) (Table 1), as previously described (4). Briefly, the real-time PCR, done using IQ SYBR green supermix (Bio-Rad), started with an initial step at 95°C for 3 min, followed by 40 cycles of 10 s at 95°C and 45 s at 63°C. Data were acquired in the final step at 63°C. Using the same genomic DNA from each sample, real-time PCRs were completed using group-specific primers to determine the amount of bacteria in each of the following major groups: the Eubacterium rectale-Clostridium coccoides group, Lactobacillus sp., Bacteroides sp., mouse intestinal Bacteroides (MIB), SFB, Enterobacteriaceae, Salmonella enterica, and total bacteria (eubacteria) (Table 1). Bacterial numbers were determined using standard curves constructed with reference bacteria specific for each bacterial group analyzed (Table 1). qPCR measures 16S gene copies per sample and not actual bacterial numbers or CFU. Nevertheless, these values are directly related and correlate well (4).

Bacterial RNA isolation and reverse transcription (RT)-qPCR.

Fresh intestinal samples were isolated from mice and snap-frozen in liquid nitrogen. Bacterial RNA was isolated using the RNeasy Plus RNA isolation kit (Qiagen). First-strand synthesis was performed using an iScript cDNA synthesis kit (Bio-Rad). The resulting cDNA was then analyzed using the qPCR procedures described above.

Statistical analysis of data.

Analyses of changes in total bacterial number and total Salmonella colonization were performed using a two-way analysis of variance, with a P value of <0.05 being considered significant. Analyses of Salmonella invasion experiments were performed using a Wilcoxon rank-sum test followed by a Tukey's multiple-comparison test, with a P value of <0.05 being considered significant. Analysis of histological scoring of enteritis was performed using a Student's t test, with a P value of <0.05 being considered significant. Analysis of changes of specific bacterial groups in response to antibiotic treatment was done as follows. For each organ, a global mixed-effects model was fitted to the log counts; the response was allowed to vary by bacterial class and treatment group, and the within-animal correlation was modeled by an unstructured correlation matrix. A Bonferroni correction was used within each organ to control for multiple testing. The pairwise comparisons of treatments within each bacterial class were protected by overall F tests.

Antibiotic treatment alters but does not eliminate the intestinal microbiota.

Antibiotic regimens (streptomycin, streptomycin-bacitracin, and AVNM) were selected because of their activities against bacterial groups common to the intestinal microbiota and their frequent use in animal models of intestinal infection and inflammation. Aerobic and anaerobic culturings of feces after antibiotic treatment showed parallel results, with significant numbers of culturable aerobes and anaerobes remaining in the feces of mice treated with streptomycin. However, the combination of streptomycin with bacitracin or the AVNM regimen resulted in the complete or near-complete elimination of culturable bacteria (not shown). Gross examination of the GI tract revealed characteristic findings, with notable swelling of the cecum ascribed to an accumulation of undigested mucus secondary to the reduction in levels of colonizing bacteria (not shown) (5). The impact of antibiotic treatment on DSI bacteria was minimal, and only the AVNM regimen resulted in significant losses represented by less than 1 log of total bacteria. The effect of antibiotics on the cecum and LI was more profound and varied depending on the specific antibiotic regimen (Fig. 1A), with total bacterial losses ranging from 1 to 3 logs. Quantitative studies using qPCR demonstrated that streptomycin had the most modest effects on the total biota (1-log loss), while the combination of streptomycin-bacitracin or AVNM had a similarly greater effect (Fig. 1A).

Alterations in microbiota composition are antibiotic and site specific.

Not only do variations in antibiotic treatment result in distinct changes in total bacterial numbers, with losses ranging from 0.5 to 3 logs of bacteria, the different antibiotic regimens also had distinct effects on the dominant bacterial groups present in the intestinal tract. Characterization of the intestinal microbiome in control mice demonstrates that the microbiota composition is site dependent with respect to the abundance of dominant bacterial groups. The colonization of the small intestine differs significantly from that of the cecum and distal colon. While present at a much lower abundance in the DSI, the most prominent bacterial group present in the cecum and LI is the E. rectale-C. coccoides group. The E. rectale-C. coccoides group is a member of the firmicute class of bacteria, and in the cecum and LI, it outnumbers all the other intestinal bacterial groups combined by 10- to 100-fold (Fig. 1B to D). All antibiotics tested had significant effects on this bacterial group in all segments of the GI tract, and losses of this group account for the majority of bacteria eliminated by antibiotics. The antibiotic effects on specific groups were consistent throughout the gut (Fig. 1B to D). For example, streptomycin-bacitracin had profound effects on the E. rectale-C. coccoides group, SFB, and Bacteroides in all segments of the gut, while streptomycin alone had a minimal impact on Bacteroides bacteria in any part of the gut. In addition to the E. rectale-C. coccoides group, SFB bacteria were also notably sensitive to antibiotic treatment. The SFB bacterial group was previously identified in a variety of animal hosts including mice (35), rabbits (15), and chickens (22). In the small intestine, SFB can be found to be in direct contact with the epithelium, where it has been ascribed a protective role in the prevention of enteric infection (15). While SFB are present and adherent to the DSI epithelium in untreated mice, all of the antibiotic regimens used eliminated SFB from direct contact with the DSI epithelium (Fig. 1E).

Incomplete recovery of the intestinal microbiome after antibiotic treatment.

The combination of streptomycin and bacitracin, while not absorbed systemically, is effective at eliminating a large percentage of the bacteria colonizing the intestinal tract. Mice were treated with this antibiotic regimen, and the recovery of their intestinal microbiome was monitored over 21 days by aerobic and anaerobic culture and qPCR. As noted above, prior to the withdrawal of antibiotics (day 0), neither aerobes (Fig. 2A) nor anaerobes (Fig. 2B) could be cultured from the LIs.

Despite the elimination of culturable bacteria, the antibiotic treatments did not eliminate bacteria, as determined by quantification by molecular methods (Fig. 1A and 3A). In addition, the abundance of culturable bacteria from the LIs of untreated mice (Fig. 2A and B) is at least 1 log less than that measured by qPCR. The discordance between the quantity of culturable bacteria and that measured by molecular methods is consistent with previously reported observations, which have demonstrated that the majority of bacteria present in the GI tract are not culturable by available methods (40). Other aspects that must be considered include the fact that qPCR measures 16S gene copy numbers, which vary between bacterial species. Previous work suggested that qPCR measurements and culture measurements correlate well when individual species are measured (4). One additional concern is whether the difference is due to picking up dead bacteria by 16S qPCR. To address this concern, bacterial RNA was isolated from selected parallel samples and analyzed by RT-qPCR. Bacterial RNA is less stable than bacterial DNA and provides a good measure of live bacteria. The total bacterial abundance was unchanged when measured by this approach (Fig. 2C).

Aerobic bacteria could not be detected 1 day after withdrawal of antibiotics but showed dramatic increases by 3 days, overgrowing the normal aerobic balance by 3 logs (Fig. 2A). This aerobic overgrowth reversed slowly and ultimately dropped slightly below levels seen in untreated mice after 3 weeks. The anaerobic bacteria showed some recovery 1 day after antibiotic withdrawal and then slightly exceeded the numbers found in untreated mice but not significantly (Fig. 2B). By 3 days after antibiotic withdrawal, the numbers of anaerobic bacteria found in antibiotic-treated and control mice were not statistically different.

Total bacterial numbers determined by qPCR were also significantly decreased compared to those for controls at 1 and 3 days after withdrawal of antibiotics (Fig. 3A). After 7 days, total numbers of bacteria had recovered (Fig. 3A), in parallel with the recovery of the dominant E. rectale-C. coccoides bacterial group (Fig. 3B). Unlike the E. rectale-C. coccoides group, the Enterobacteriaceae are present in very low numbers in the murine gut. This group expands within 7 days after antibiotic treatment and persists at elevated levels, but by 3 weeks, it is reduced back to baseline levels (Fig. 3C). Other bacterial groups examined do not show complete recovery as quickly. Bacteroides groups are highly variable during recovery from antibiotics (not shown); however, SFB levels stay consistently low and had not recovered after 3 weeks (Fig. 3D). Although the abundance and distribution of bacterial groups vary between the DSI and the rest of the lower GI tract, the recovery of specific bacterial groups is temporally consistent throughout each section of the intestine. For example, levels of the Enterobacteriaceae peak on day 7 (Fig. 3C and not shown) and return to baseline levels by day 21 in all segments of the GI tract. Overall, despite the fact that recovery is not complete after 3 weeks, the intestinal microbiota composition gravitated back toward that of untreated mice over time.

Salmonella invasion correlated with microbial abundance and recovery of the E. rectale-C. coccoides population.

To determine whether the extent of changes in the microbiota increased host susceptibility to bacterial translocation, mice treated with either streptomycin, streptomycin-bacitracin, or AVNM were challenged with Salmonella. Salmonella burden in the livers and spleens of infected mice was analyzed. Mice not treated with antibiotics showed low-level Salmonella translocation into the spleen and liver 3 days postinfection (Fig. 4A), while pretreatment with antibiotic regimens resulted in significantly increased levels of invasion of Salmonella from the lumen to the systemic circulation, as shown previously with streptomycin treatment (30). The extent of invasion varied between antibiotic treatments but not significantly. Surprisingly, pretreatment with streptomycin-bacitracin or AVNM, although effective at eliminating large numbers of colonizing bacteria, did not promote significantly more Salmonella invasion than did streptomycin treatment alone (Fig. 4A). While this suggests that the absolute quantity of biota loss did not correlate with susceptibility to pathogen translocation, it is possible that there is a dose-dependent relationship between bacterial abundance and translocation (36) but that even the mildest antibiotic treatment used in this study was too high a dose to detect dose dependence. Correlation is evident between Salmonella translocation and the loss of several firmicute groups in every section of the GI tract, including the E. rectale-C. coccoides group, Lactobacillus sp., and SFB. Following the above-described time course for the recovery of the intestinal microbiome, antibiotics were withdrawn; groups of mice were allowed to recover for 1, 3, 7, 14, and 21 days before being orally challenged with Salmonella; and indices of infection were compared to those of non-antibiotic-treated control groups. Increased susceptibility to pathogen translocation disappeared by 7 days after antibiotic withdrawal (Fig. 4B), in direct relation to the recovery of both the E. rectale-C. coccoides (Fig. 3B) and Lactobacillus (not shown) groups. Since the recovery of the E. rectale-C. coccoides group resulted in the recovery of total bacterial numbers, it is also possible that the restoration of total bacterial abundance rather than the specific bacterial group is the critical factor in the prevention of bacterial translocation.

Antibiotic disruption of the intestinal microbiota results in increased Salmonella colonization.

After antibiotic treatment, mice were given Salmonella by oral gavage, and the intestinal microbiota was examined for Salmonella abundance in the intestinal lumen (Fig. 5A). The level of Salmonella colonization of control mice was relatively low, consistent with previously reported findings (4). Antibiotic treatment of the mice resulted in increased levels of Salmonella colonization compared to that for control mice (Fig. 5A). Although the change in total bacterial quantity varied significantly between antibiotic treatments, the intestinal Salmonella burden was consistent between treatment groups, at approximately 10⁶ 16S Salmonella gene copies per gram of intestine in the DSI and 10⁹ 16S Salmonella gene copies per gram in the cecum and LI, demonstrating that there is no direct relationship between the extent of biota loss and the abundance of Salmonella colonization. As previously noted (4), during an enteric infection, Salmonella does not outcompete or replace the native biota. This suggests that the change in the total bacterial numbers may not be the most important factor in Salmonella colonization, but it is also possible that the antibiotic treatments used exceeded the range where this could be determined. To distinguish between these possibilities, intestinal colonization by Salmonella was examined with mice challenged during recovery from antibiotics. Three weeks after antibiotic withdrawal, several of the parameters of the intestinal microbiome showed recovery. This recovery included recovery of total bacterial numbers, several dominant bacterial groups, and the anaerobic population and near recovery of the aerobic population. Nevertheless, animals challenged orally with Salmonella remained susceptible to increased pathogen colonization of the intestinal tract (Fig. 5B) and again appeared to colonize to a fixed set point of 10⁹ 16S gene copies per gram in the cecum and LI, representing approximately 1% of the total bacterial population at these sites.

Alteration of the intestinal biota results in increased intestinal mucosal inflammation.

Previous work by our laboratory (4) demonstrated that Salmonella infection of mice not treated with antibiotics results in acute inflammation of the DSI. Work by other laboratories has shown that streptomycin treatment followed by Salmonella infection results in profound mucosal inflammation in the ceca of mice (5). To determine the effect of microbiota disruption on the development of intestinal inflammation, the intestinal tissues of antibiotic-treated Salmonella-infected mice were analyzed. In control mice, low-level focal inflammation was noted in the DSI (4), cecum, and LI (Fig. 6). Antibiotic treatment increased inflammation minimally in the DSI, but significantly more inflammation was observed in the cecum and LI (Fig. 6). This is consistent with previous work by Barthel et al. in the characterization of a streptomycin pretreatment mouse model for Salmonella colitis (5). Cecal inflammation involved the entire mucosal surface and was characterized by edema, extensive neutrophil influx into the lamina propria and the lumen with evidence of crypt abscesses and crypt destruction, and mucosal erosion and ulceration (Fig. 6). Again, although the antibiotics had diverse effects on the intestinal biota, the extent of these changes was not directly related to the extent of inflammation. There were minimal differences between the levels of inflammation in streptomycin-treated mice (mild changes in the biota) and those in AVNM-treated mice (profound changes in the biota). Again, this finding was confirmed by antibiotic recovery experiments. Similar to the colonization results, the extent of cecal inflammation in response to Salmonella infection remains statistically greater in mice treated by antibiotics than in non-antibiotic-treated controls even when mice are allowed to recover from antibiotic treatment for 3 weeks before oral Salmonella infection (Fig. 7).