The effect of antibiotics and bismuth on fecal hydrogen sulfide and sulfate-reducing bacteria in the rat

Abstract

Colonic bacteria produce the highly toxic thiol, hydrogen sulfide. Despite speculation that this compound induces colonic mucosal injury, there is little information concerning manipulations that might reduce its production. We studied the effect of antibiotics and bismuth on the production of hydrogen sulfide in rats. Baseline fecal samples were analyzed for hydrogen sulfide concentration and release rate during incubation and numbers of sulfate-reducing bacteria. Groups of six rats received daily doses of ciprofloxacin, metronidazole, or sulfasalazine for one week, and feces were reanalyzed. Bismuth subnitrate was then added to the antibiotic regimens. While sulfide production and sulfate-reducing bacteria were resistant to treatment with ciprofloxacin or metronidazole, bismuth acted synergistically with ciprofloxacin to inhibit sulfate-reducing bacteria growth and to reduce sulfide production. Combination antibiotic–bismuth therapy could provide insights into the importance of sulfide and sulfate-reducing bacteria in both human and animal models of colitis and have clinical utility in the treatment of antibiotic-resistant enteric pathogens.

1 Introduction

Colonic bacteria produce large quantities of hydrogen sulfide (H₂S). Like cyanide, this volatile thiol binds to the heme moiety of cytochrome a₃, blocking the terminal step in electron transport [1]. When administered systemically, H₂S has a median lethal dose (LD₅₀) for rodents similar to that of cyanide [2].

The volatility of H₂S results in its rapid dissociation from fecal material, and this gas has very high tissue permeability [3]. Thus, the colonocyte is exposed to virtually the entire bacterial output of this compound, in contrast to other potentially toxic bacterial metabolites, which remain in feces or which have low permeability for the colonic mucosa. While the colonocyte efficiently detoxifies H₂S via conversion to thiosulfate [4], there has been speculation that this highly toxic compound could play a pathogenetic role in colonic diseases, particularly ulcerative colitis (UC) [5]. Sulfide also has been linked to colon neoplasia via the observation that sulfide exposure induced proliferation in the upper crypt region of the colonic mucosa [6], a finding associated with mucosal hyperplasia.

There is little information concerning factors that influence colonic production of H₂S. It is known that H₂S is produced by a group of bacteria, sulfate-reducing bacteria (SRB), that obtain energy via reactions involving the reduction of sulfate to sulfide [7]. However, in vitro studies indicate that fecal bacteria liberate H₂S far more efficiently from organic sulfur-containing compounds than from sulfate [8]. Thus, it is not clear if SRB metabolism determines the rate of H₂S production in the gut.

Understanding the role of SRB in colonic conditions would be enhanced by the availability of a manipulation that inhibits SRB and/or reduces the production of H₂S. In the present study, we studied the influence of four compounds, metronidazole, ciprofloxacin, sulfasalazine and bismuth subnitrate (BSN) on SRB counts and production of H₂S in the intestinal tract of rats.

2 Materials and methods

2.1 Animals

Groups of six male Sprague–Dawley rats (weight 300–440 g) were studied. The basal diet was rat chow (Harlan Teklad, Madison, WI, USA), ground to powder to permit the admixture of test compounds. The diets were consumed ad lib, and food ingestion and weight of the rats was monitored.

2.2 Treatments and sample collection

After one week on chow, rats were induced to provide fresh fecal samples via palpation of the lower abdomen. The fecal samples passed directly from the anus into pre-weighed containers: 20-ml polypropylene syringes for measurement of H₂S release; pre-weighed 50-ml test tubes containing 5 ml of 2% zinc acetate (ice cold) for H₂S concentration measurements; and 4-ml vials for SRB culture. One group continued to ingest the untreated chow (controls) while the other groups received either metronidazole, ciprofloxacin, or sulfasalazine added to chow in concentrations such that if the rat ingested the expected 25 g of chow daily, the approximate daily dosages would be: metronidazole 30 mg, ciprofloxacin 20 mg, or sulfasalazine 80 mg. The three doses would be roughly four times the human dose per unit body weight. After 7 days, fecal samples were again collected as described, and BSN was added to the chow of all four groups in a concentration of 6.5%. After 7 days, fecal samples were collected, BSN was discontinued while antibiotic treatment was maintained, and fecal collections were obtained at multiple points over a two-week period.

2.3 H₂S release by incubated fecal specimens

Following collection of the fecal sample, the plunger was immediately reinserted into the syringe, the gas space was flushed with nitrogen (N₂), 20 ml of N₂ was added, and the syringe was sealed with a stopcock. The syringes were incubated at 37°C, and at 1, 2, 4, and 24 h the gas volume was recorded and a 0.3-ml gas sample was removed for analysis for H₂S. At 24 h, the syringes were weighed, and fecal weight determined by difference. The rate of release of H₂S per gram wet weight was calculated from the concentration of H₂S, the volume of the gas space, and fecal weight.

2.4 H₂S concentration of feces

The vial containing the fecal sample and zinc acetate was weighed and fecal weight determined by difference. The volume of zinc acetate was adjusted to give a 1:100 ratio of feces to zinc acetate and the sample was homogenized. A 0.5-ml aliquot of the homogenate was added to a 20-ml polypropylene syringe, 0.5 ml of 12 N HCl was instilled into the syringe, and the syringe was sealed with a stopcock. After 30 min, a 0.3-ml aliquot of gas space was analyzed for H₂S.

2.5 Bacteriology

Viable SRB were quantitated using a method described by Jain [9]. Postgate's B medium, preinoculated with non-SRB bacteria (Hafnia alvei), was incubated to create anaerobic conditions as evidenced by a change of the indicator, resazurin, from pink to colorless. Serial 10-fold dilutions of fecal homogenates made in phosphate buffer (pH 7.0) were instilled into the pre-reduced media and incubated for 14 days at 37°C. Growth of SRB was evidenced by black staining (ferrous sulfide) of the media. The highest dilution showing growth (log₁₀ colony-forming units (CFU) per gram wet weight stool) was recorded and used in calculations. However, the true counts would fall between this value and the log of the next highest dilution.

2.6 Statistical analysis

All data were expressed as mean±standard error of the mean (S.E.M.). Characteristics of different treatment groups were compared with the unpaired t-test. Paired t-tests were used to examine changes over the trial period within each group. Probability values of P<0.05 were taken as significant.

3 Results

3.1 Influence of antibiotics on fecal H₂S concentration, release, and SRB counts

Fecal sulfide concentration, which averaged 0.93±0.20 µmol g⁻¹ prior to antibiotic treatment, was not significantly influenced by metronidazole or sulfasalazine but was reduced (P<0.05) by ciprofloxacin (Fig. 1). Administration of metronidazole, ciprofloxacin or sulfasalazine had no statistically significant effect on H₂S release by feces at any time point during incubation (Fig. 2). None of the antibiotics (in the absence of BSN) significantly altered SRB counts (Fig. 3).

3.2 Influence of antibiotics plus BSN on fecal H₂S concentration, release and SRB counts

The feces of non-antibiotic-treated rats receiving BSN were black and released negligible H₂S upon incubation at 37°C. The sulfide concentration of these feces (mean: 23.2±4.5 µmol g⁻¹) was similar to that of rats receiving sulfasalazine plus BSN (24.8±4.3 µmol g⁻¹). In contrast, the feces of four of the six ciprofloxacin–BSN-treated rats were brown and had negligible fecal sulfide concentrations (mean: 0.2±0.1 µmol g⁻¹) while fecal sulfide of the other two rats approached that of BSN-treated controls. The feces of five of six rats receiving metronidazole and BSN also were brown and had sulfide concentrations below the range observed in BSN controls, while one rat had fecal sulfide similar to the controls. The mean values for the six rats receiving bismuth plus either metronidazole or ciprofloxacin were significantly less (P<0.05) than that of the BSN controls (Fig. 4). Following removal of BSN from the diets of the metronidazole-treated rats, the low H₂S release returned to normal levels within 2 days, whereas 15 days were required for normalization of H₂S release in the ciprofloxacin-treated animals (Fig. 5).

As shown in Fig. 3, SRB counts of untreated rats (9.2±1.33 log₁₀ counts g⁻¹) did not differ significantly from those treated with BSN alone (9.3±0.52 log₁₀ counts g⁻¹) or the combination of BSN plus metronidazole (8.3±0.41 log₁₀ counts g⁻¹) or BSN plus sulfasalazine (8.3±0.41 log₁₀ counts g⁻¹). The four ciprofloxacin–bismuth-treated animals whose fecal H₂S concentration was negligible, had roughly five log₁₀ reductions in SRB counts (see Fig. 3), whereas the two rats with near normal H₂S release showed no decline in SRB. The average SRB count of all six rats was significantly changed from 8.3±0.52 for ciprofloxacin alone to 5.8±2.86 log₁₀ counts g⁻¹ of feces when BSN was combined with ciprofloxacin (P<0.05).

4 Discussion

Despite speculation that H₂S could play an etiological role in colonic disease, there are limited data concerning manipulations that might reduce the production of this compound in the human or animal colon. In particular, we are aware of no data concerning the influence of antibiotic administration on the in vivo production of H₂S.

Most publications purporting to assess colonic sulfide production have measured fecal sulfide concentration [5,7,10–12]. Sulfide exists in ionized, non-volatile states (S⁻ or HS⁻), as well as volatile H₂S [13]. While feces contain measurable concentrations of sulfide, our recent studies indicated that about 95% of the sulfide produced in the gut is absorbed and the vast majority of the 5% passed in feces is tightly bound to other compounds [14]. Thus, fecal sulfide concentration appears to reflect sulfide binding capacity rather than intraluminal production.

In addition to fecal concentration measurements, we employed three other techniques to assess the influence of antibiotics on H₂S production. Fecal samples were anaerobically incubated at 37°C, and the release of H₂S was quantitated, a measurement that might not accurately reflect production in the proximal colon. Therefore, we also measured the total gut production of sulfide via the oral administration of bismuth as BSN, which binds all sulfide produced in the gut. Measurement of sulfide release following fecal acidification provides a quantitative assessment of total sulfide production. Lastly, fecal SRB were quantitated as described by Jain [9], which utilizes Postgate's B medium that is pre-reduced via incubation with a non-sulfide-producing, oxygen-consuming organism (H. alvei), a manipulation that increases SRB counts by 2- to 3-fold log₁₀ over that with untreated media.

Incubated feces of control rats released H₂S at a rate of 0.17±0.02 µmol g⁻¹ wet weight h⁻¹, a value roughly similar to that reported with human feces (0.088 µmol g⁻¹ wet weight h⁻¹) [8]. The fecal concentration of SRB in control animals (log₁₀ CFU g⁻¹ wet weight) ranged from >10⁸ to >10¹¹. These values are higher than those reported for 16 healthy human subjects, the majority of whom had <10⁸ SRB g⁻¹ fecal dry weight [12], a discrepancy possibility attributable to our use of pre-reduced Postgate's B media in the present study.

None of the antibiotics tested significantly influenced either the release of H₂S by incubated fecal samples (Fig. 2) or the fecal SRB counts (Fig. 3). Thus, SRB (or other bacteria responsible for H₂S release) in the distal colon of rats are resistant, in vivo, to individual antibiotics with broad activity against anaerobes (metronidazole) and aerobes (ciprofloxacin) as well as a compound (sulfasalazine) reported to reduce sulfide production by human fecal homogenates [15]. A previous report found that Desulfovibrio, the SRB most commonly isolated from human feces, was resistant to a variety of antibiotics [16]. Inhibition was observed only with gentamicin at a concentration of 25 µg ml⁻¹, a concentration greater than the 8 µg ml⁻¹ commonly considered to represent the upper limit of sensitivity.

After obtaining the above measurements, BSN was administered to measure total sulfide release in the gut. The fecal sulfide concentration of the non-antibiotic-treated animals averaged 23.2±4.5 µmol g⁻¹ feces, a value roughly 24 times greater than that (0.93±0.2 µmol g⁻¹) observed in non-BSN-treated rats. Given this 24-fold discrepancy in sulfide concentration, it appears that in the absence of BSN about 95% (23/24) of colonic sulfide escaped from the fecal stream and was absorbed.

While sulfasalazine had no discernible effect on fecal sulfide concentration in BSN-treated animals, co-administration of metronidazole or ciprofloxacin and BSN resulted in markedly reduced sulfide concentrations relative to that observed with BSN alone. For example, four of six rats receiving ciprofloxacin plus BSN had negligible fecal sulfide (see Fig. 4), and the metronidazole plus BSN regimen resulted in approximately 5-fold reductions in fecal sulfide in five of six animals (see Fig. 4). The brown stools passed by animals with markedly decreased fecal sulfide, in contrast to the black stools of the BSN controls, provided visual confirmation of the chemical measurement. A previous study showed that addition of BSN to fecal homogenates [17] did not perturb sulfide production. Thus, it appears that BSN works synergistically with ciprofloxacin or metronidazole in vivo to reduce sulfide production in the colon of some but not all animals.

The four rats with negligible gut sulfide production during ciprofloxacin plus BSN therapy had roughly five log₁₀ decreases in SRB while SRB remained normal in the two animals with minor decreases in fecal sulfide. This strong correlation suggests that SRB play a crucial role in colonic H₂S production. However, decreased sulfide production of metronidazole–BSN-treated rats was not associated with a reduction in SRB suggesting that this combination resulted in in situ inhibition of SRB metabolism but not SRB replication. This hypothesis is supported by the observation that following removal of BSN from BSN–metronidazole regimen, fecal sulfide production returned to normal within 2 days, whereas 14 days were required for normalization following withdrawal of BSN from the BSN–ciprofloxacin regimen.

The major new observation of the present study is that while sulfide production was not influenced by administration of several different antibiotics, this production was inhibited by the combination of BSN and either ciprofloxacin or metronidazole. Although the antibacterial mechanism of bismuth is not well understood, bismuth preparations have been widely used for the treatment of Helicobacter pylori[18,19] and as prophylaxis for traveler's diarrhea [20]. Investigation of the ability of orally administered bismuth to alter the colonic flora appears to be limited to a single human study in which 8 ounces of bismuth subsalicylate administered daily for 2 days had no demonstrable effect on total fecal microbial counts or counts of Pseudomonas spp., Staphylococcus spp., Bacteroides spp., or Clostridium difficile[21]. We are aware of no data concerning the effect of antibiotic–bismuth combination therapy on the human colonic flora. The observation that BSN acted synergistically with ciprofloxacin to inhibit SRB growth raises the possibility that, similar to the situation in H. pylori therapy, co-administration of bismuth might enhance the activity of antibiotics against a variety of enteric organisms.

The clinical importance of the observation that antibiotic–BSN combinations can reduce fecal sulfide and SRB counts hinges on whether SRB or their metabolic product, sulfide, are capable of inducing tissue injury. Several animal models of colitis employ feeding of non-absorbable forms of sulfate (e.g. dextran sulfate and carageenan). Since bacterial activity is necessary for toxicity, some bacterial metabolite of these compounds apparently is the injurious agent. However, the relevance of these experimental models to human disease remains highly speculative. More compelling are the studies of Roediger and colleagues [6], who demonstrated that experimental exposure of colonic mucosa to sulfide inhibited butyrate utilization, a defect similar to that observed in the mucosa of subjects with UC.

The failure of antibiotic administration to bring about remission in UC [22,23] has been interpreted to indicate that the colonic flora does not play a major pathogenetic role in this condition. However, our observations in the rat suggest that SRB are extremely resistant to most antibiotics; thus, it is not clear that previous use of antibiotics in UC subjects actually influenced fecal SRB counts or sulfide production. The finding that BSN plus ciprofloxacin can inhibit SRB growth and fecal sulfide release may provide a means of testing the concept that SRB play a role in colitis in humans or in animal models involving the administration of non-absorbable sulfur-containing compounds.

Acknowledgements

This study was partly supported by ‘The Japan Antibiotics Research Association — Pfizer Infectious Diseases Research Fund 2002’.

References