- Split View
-
Views
-
Cite
Cite
Jane Freeman, Simon D. Baines, Katie Saxton, Mark H. Wilcox, Effect of metronidazole on growth and toxin production by epidemic Clostridium difficile PCR ribotypes 001 and 027 in a human gut model, Journal of Antimicrobial Chemotherapy, Volume 60, Issue 1, July 2007, Pages 83–91, https://doi.org/10.1093/jac/dkm113
- Share Icon Share
Abstract
We compared the behaviour of Clostridium difficile PCR ribotypes 001 and 027 in a human gut model, and compared the responses to metronidazole exposure.
Using a human gut model primed with pooled human faeces, gut flora bacterial counts, C. difficile total viable counts, spore counts and cytotoxin titres were determined, following exposure to clindamycin, in the absence or presence of metronidazole.
Duration of cytotoxin production by C. difficile ribotype 027 was markedly longer than that of ribotype 001 (23 versus 13 days, respectively), but peak toxin titres were similar. During toxin production, total C. difficile ribotype 027 populations had higher proportions of vegetative cells than did ribotype 001 (median 56.33 versus 23.54%). Similarly, total C. difficile ribotype 027 populations remained predominantly as vegetative cells for longer than did ribotype 001 (20 versus 9 days). The effects of metronidazole on C. difficile were markedly less than expected. Titres of C. difficile ribotype 001 cytotoxin were reduced but recurred following metronidazole administration. C. difficile ribotype 027 cytotoxin titres in the distal section of the gut model were unaffected by metronidazole. These observations correlated with poor metronidazole concentrations.
Duration of cytotoxin production by C. difficile ribotype 027 markedly exceeds that of ribotype 001. Sub-optimal gut concentrations of metronidazole, possibly due to inactivation by components of normal gut flora, are associated with continued toxin production. These findings may help to explain the increased severity of symptoms and higher case-fatality ratio associated with infections due to C. difficile ribotype 027.
Introduction
Clostridium difficile is implicated in 20–30% of cases of antibiotic-associated colitis and is the aetiological agent of pseudomembranous colitis (PMC).1,2 Most antimicrobials have been implicated, but clindamycin, third-generation cephalosporins and aminopenicillins are particularly noted for their propensity to induce C. difficile infection (CDI).3–5 Reporting of C. difficile in England is now mandatory as part of a national healthcare-associated infection surveillance initiative.6 Laboratory reports of C. difficile in England and Wales increased more than 15-fold during the 1990s,7 and by 23% between 2003 and 2004.8 Nosocomial CDI is a significant burden on healthcare services and results in additional costs associated with prolonged hospital stay, intensive nursing, isolation, laboratory and infection control facilities. In 1996, the additional costs due to CDI were over £4000 per case,9 while a more recent US study estimated healthcare costs associated with C. difficile diarrhoea at over $1.1 billion.10
Epidemiological studies of C. difficile have reported the predominance of specific C. difficile strains. Stubbs et al.11 reported the presence of C. difficile PCR ribotype 001 in 33 out of 58 UK hospitals. C. difficile PCR ribotype 001 is known to be endemic in multiple countries including the US and Canada.12–14 Recently, outbreaks associated with a hitherto uncommon C. difficile strain, PCR ribotype 027 (also known as NAP1), have occurred in the US, Canada, UK and Europe.15–19C. difficile PCR ribotype 027 is thought to produce increased levels of toxins A and B, possibly as a result of an 18 bp deletion in the putative negative regulator of toxin gene expression, tcdC.16,20 Increased disease severity and case fatality, and sub-optimal efficacy of metronidazole treatment have typically been seen in C. difficile PCR ribotype 027 associated outbreaks.16–22
We have previously described an in vitro human gut model of CDI that yields results consistent with clinical and animal model data.23–26 We aimed to use this model to compare growth and toxin production by two epidemic C. difficile strains, PCR ribotypes 027 and 001, and to determine the relative efficacy of metronidazole exposure in CDI.
Material and methods
Bacterial strains
C. difficile PCR ribotypes 001 (a highly prevalent UK epidemic strain) and 027 (an emerging hyper-virulent strain) were investigated in the gut model.
Gut model
The gut model was developed and validated against the caecal contents of sudden death victims by MacFarlane et al.27 We have previously described its modification and use as a model of CDI.23–26 Briefly, the model consists of three vessels operating in a weir cascade system in an oxygen-free nitrogen atmosphere. The system is top-fed with growth medium at a controlled rate, and each vessel operates at a controlled pH to reflect the increasing alkalinity and decreasing substrate availability of the gut from proximal to distal. Thus, vessel 1 (280 mL) operates at a low pH (5.5) and high substrate availability, whereas vessels 2 and 3 (300 mL) operate at more neutral pH (6.2 and 6.8, respectively) and lower substrate availability. The model is primed with pooled emulsified faecal samples, and allowed to equilibrate in respect of bacterial populations for 2 weeks at a total retention time of 66.7 h (vessel 1 = 16.7 h, vessel 2 = 25 h, vessel 3 = 25 h).
Preparation of gut model
Human faeces were collected from five healthy elderly volunteers (>65 years) with no history in the previous 2 months of antibiotic treatment. Faecal samples were transported under anaerobic conditions to the laboratory (GasPak, Oxoid, Basingstoke, UK), and were confirmed as C. difficile-negative.23–26 A 10% (w/v) coarse-filtered faecal slurry was prepared and used to inoculate the gut model. Gut model growth medium was prepared as previously described and resazurin anaerobic indicator (0.005 g/L) was added through a sterile filtration device.23–26 The media pump was started and the system left for ∼2 weeks to achieve steady-state bacterial populations. The model was sampled daily thereafter for enumeration of C. difficile total and spore counts, toxin titres and for quantification of gut bacterial populations.
Effect of clindamycin exposure on C. difficile PCR ribotypes 001 and 027 in the gut model
Clindamycin-induced toxin production was achieved as previously described.25 Briefly, at steady-state, vessel 1 of the gut model was inoculated with ∼107C. difficile spores. There were no further interventions for 7 days. The gut model was then inoculated with a further 107C. difficile spores, and clindamycin was instilled into vessel 1 at 33.9 mg/L, 6 hourly, to maintain the biliary/faecal levels observed following a single 600 mg dose.28 After 7 days, clindamycin instillation ceased and no further interventions were made. The experimental endpoint was defined as a decrease in C. difficile toxin titres to ≤1 relative unit (RU) accompanied by C. difficile sporulation. The experimental time-course is represented in Table 1.
. | Time period . | ||||
---|---|---|---|---|---|
. | A . | B . | C . | D . | E . |
Duration (days) | 14 | 7 | 7 | 10a | 14 |
Description | Steady-state | C. difficile spore instillation | C. difficile spore instillation+33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily |
. | Time period . | ||||
---|---|---|---|---|---|
. | A . | B . | C . | D . | E . |
Duration (days) | 14 | 7 | 7 | 10a | 14 |
Description | Steady-state | C. difficile spore instillation | C. difficile spore instillation+33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily |
aTime taken to induce toxin production by C. difficile was variable.
. | Time period . | ||||
---|---|---|---|---|---|
. | A . | B . | C . | D . | E . |
Duration (days) | 14 | 7 | 7 | 10a | 14 |
Description | Steady-state | C. difficile spore instillation | C. difficile spore instillation+33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily |
. | Time period . | ||||
---|---|---|---|---|---|
. | A . | B . | C . | D . | E . |
Duration (days) | 14 | 7 | 7 | 10a | 14 |
Description | Steady-state | C. difficile spore instillation | C. difficile spore instillation+33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily |
aTime taken to induce toxin production by C. difficile was variable.
Effects of metronidazole on C. difficile PCR ribotypes 001 and 027 following induction of toxin production by clindamycin
Clindamycin-induced high-level toxin production was achieved as previously described.25 At steady-state, the gut model was inoculated with ∼107C. difficile spores. There were no further interventions for 7 days. The gut model was then inoculated with a further 107C. difficile spores, and clindamycin was instilled into vessel 1 at 33.9 mg/L, 6 hourly, to maintain the biliary/faecal levels observed following a single 600 mg dose.28 After 7 days, clindamycin instillation ceased and no further interventions were made. When a high C. difficile toxin titre was achieved and maintained for 2 consecutive days, metronidazole instillation commenced into vessel 1 (9.3 mg/L 8 hourly for 7 days to reflect biliary/faecal levels achieved).29 Following cessation of metronidazole treatment, there were no further interventions for 2 weeks. The experimental time-course is represented in Table 2.
. | Time period . | |||||
---|---|---|---|---|---|---|
. | A . | B . | C . | D . | E . | F . |
Duration (days) | 14 | 7 | 7 | 10a | 7 | 14 |
Description | steady-state | C. difficile spore instillation | C. difficile spore instillation +33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | 9.3 mg/L MET 8-hourly once toxin induced | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily | Daily |
. | Time period . | |||||
---|---|---|---|---|---|---|
. | A . | B . | C . | D . | E . | F . |
Duration (days) | 14 | 7 | 7 | 10a | 7 | 14 |
Description | steady-state | C. difficile spore instillation | C. difficile spore instillation +33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | 9.3 mg/L MET 8-hourly once toxin induced | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily | Daily |
MET, metronidazole.
aTime taken to induce toxin production by C. difficile was variable.
. | Time period . | |||||
---|---|---|---|---|---|---|
. | A . | B . | C . | D . | E . | F . |
Duration (days) | 14 | 7 | 7 | 10a | 7 | 14 |
Description | steady-state | C. difficile spore instillation | C. difficile spore instillation +33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | 9.3 mg/L MET 8-hourly once toxin induced | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily | Daily |
. | Time period . | |||||
---|---|---|---|---|---|---|
. | A . | B . | C . | D . | E . | F . |
Duration (days) | 14 | 7 | 7 | 10a | 7 | 14 |
Description | steady-state | C. difficile spore instillation | C. difficile spore instillation +33.9 mg/L clindamycin every 6 h | Post-dosing toxin induction | 9.3 mg/L MET 8-hourly once toxin induced | Recovery |
Sampling frequency | Once every 2 days | Daily | Daily | Daily | Daily | Daily |
MET, metronidazole.
aTime taken to induce toxin production by C. difficile was variable.
Enumeration of bacteria and cytotoxin measurement
C. difficile (total counts and alcohol-shock spore counts) and faecal bacteria (total facultative aerobes, total anaerobes, lactose fermenters, enterococci, lactobacilli, bacteroides, clostridia and bifidobacteria) were enumerated by viable counting of serial 10-fold dilutions in peptone water on selective media.23–27 Colonies were identified by colony morphology, colony fluorescence, Gram stain, biochemical reactivity, antimicrobial susceptibility and genus-specific PCR where appropriate. C. difficile cytotoxin was quantified in a VERO cell culture cytotoxicity assay with Clostridium sordellii neutralization, as previously described.23–26
Assay of clindamycin and metronidazole
Active clindamycin concentrations were determined by large plate bioassay using Kocuria rhizophila ATCC 9341 as an indicator organism.30 Metronidazole concentrations were determined by in-house large-plate multi-well diffusion assay. Briefly, 100 mL volumes of Columbia blood agar (Oxoid) were poured into 245 mm2 assay plates (NUNC, Roskilde, Denmark). A laboratory strain of Clostridium sporogenes (indicator organism) was cultured anaerobically for 24 h on pre-reduced FBA, and used to inoculate pre-reduced Schaedler's anaerobic broth. Following 24 h anaerobic incubation at 37°C, the culture was mixed, the assay medium surface flooded, excess liquid culture removed and plates allowed to dry in an anaerobic environment. Twenty-five wells (9 mm diameter) were cut in each agar plate using a cork borer and the wells randomly filled with one drop from a Pasteur pipette of either a calibrator or a test sample from the gut model. Calibrators consisted of doubling dilutions of metronidazole (Sigma) (0.5–256 mg/L), prepared in distilled water and sterilized by filtration. Samples from the gut model were centrifuged at 16 000 g for 10 min and the supernatants sterilized by filtration. Inoculated plates were incubated anaerobically for 24 h. Zone diameters (mm) were measured and a calibration line produced by plotting the square of the zone diameter against the log2 concentration of metronidazole. All assays were performed in duplicate. The log2 metronidazole concentrations in the gut model samples were read from the calibration line and converted into the actual concentrations using a 2x conversion. Assay sensitivity limit was 0.5 mg/L.
Results
Effect of clindamycin exposure on C. difficile PCR ribotypes 001 and 027 in the gut model
Following addition to the gut model as spores, C. difficile PCR ribotype 027 remained quiescent during the next 7 days, showing no evidence of germination or toxin production (Figure 1a and b). Following commencement of clindamycin instillation and a further inoculum of C. difficile spores into vessel 1, there was no evidence of germination of C. difficile PCR ribotype 027 spores until 6 days post-cessation of antibiotic dosing (day 32), at which point C. difficile counts had decreased to <103 cfu/mL (Figure 1b, period C). Increased cytotoxin titres were observed 1 day later (day 33). While viable counts of C. difficile PCR ribotype 001 were slightly higher than those of PCR ribotype 027 at the start of period D (5 versus 4 log10 cfu/mL), comparator data for C. difficile PCR ribotype 001 were very similar up until this point. Thus, germination was observed 7 days post-cessation of clindamycin instillation (day 33), indicated by the divergence of total viable counts (TVC) and spore counts, with increased cytotoxin titres following 2 days later (day 35).
Germination of C. difficile PCR ribotype 027 was enhanced in comparison with that of C. difficile PCR ribotype 001, exhibiting rapid growth increase of over 5 log10 cfu/mL (Figure 1a and b, period D). TVC and spore counts of C. difficile ribotype 027 remained divergent for 20 days in marked comparison to those of C. difficile PCR ribotype 001 (9 days). The proportion of the TVC accounted for by vegetative cells was calculated during the germination periods of each strain (9 and 20 days, respectively). A median 24% (range 11–35%) of C. difficile PCR ribotype 001 were vegetative cells versus 56% (13–100%) of C. difficile PCR ribotype 027 during these periods. The proportions of TVC which were vegetative cells were significantly greater in C. difficile PCR ribotype 027 than ribotype 001 in both vessel 2 (P = 0.001) and vessel 3 (P = 0.008) of the gut model over the first 8 days of active germination (vessel 2, Wilcoxon signed ranks test; vessel 3, paired t-test).
Increased cytotoxin titres were observed in both vessels 2 and 3 one day after germination of C. difficile PCR ribotype 027 and coincided with the observed rapid growth (data for vessel 2 not shown). Elevated cytotoxin titres were maintained throughout the 20 day period while C. difficile PCR ribotype 027 was in the vegetative state, decreasing below 2 RU as TVC and spore counts reconverged (Figure 1b, period E). C. difficile PCR ribotype 027 cytotoxin titres were less uniform than PCR ribotype 001. The durations of toxin production by C. difficile PCR ribotypes 027 and 001 differed markedly (23 versus 13 days, respectively; Table 3). Consequently, the total measured relative cytotoxin units also differed markedly (66.5 versus 44 RU). However, there was no difference in the peak yield of cytotoxin achieved (5 RU), and indeed the average cytotoxin titres throughout the respective vegetative periods of the two strains were similar (2.9 versus 3.4 RU for C. difficile PCR ribotype 027 versus 001, respectively; Table 3). Cytotoxin titres of C. difficile PCR ribotype 027 did not decrease below an experimental endpoint of < 1 RU, but the experiment was discontinued due to time constraints.
Ribotype . | Duration TVC>SPa (days) . | Toxin produced (days post-clindamycin) . | Duration of toxin (days) . | Maximal titre (RU) . | Average titre (RU) . | Total toxin produced (RU) . |
---|---|---|---|---|---|---|
001 | 9 | 9 | 13 | 5 | 3.4 | 44 |
027 | 20 | 7 | 23 | 5 | 2.9 | 66.5 |
Ribotype . | Duration TVC>SPa (days) . | Toxin produced (days post-clindamycin) . | Duration of toxin (days) . | Maximal titre (RU) . | Average titre (RU) . | Total toxin produced (RU) . |
---|---|---|---|---|---|---|
001 | 9 | 9 | 13 | 5 | 3.4 | 44 |
027 | 20 | 7 | 23 | 5 | 2.9 | 66.5 |
aSP, spore count.
Ribotype . | Duration TVC>SPa (days) . | Toxin produced (days post-clindamycin) . | Duration of toxin (days) . | Maximal titre (RU) . | Average titre (RU) . | Total toxin produced (RU) . |
---|---|---|---|---|---|---|
001 | 9 | 9 | 13 | 5 | 3.4 | 44 |
027 | 20 | 7 | 23 | 5 | 2.9 | 66.5 |
Ribotype . | Duration TVC>SPa (days) . | Toxin produced (days post-clindamycin) . | Duration of toxin (days) . | Maximal titre (RU) . | Average titre (RU) . | Total toxin produced (RU) . |
---|---|---|---|---|---|---|
001 | 9 | 9 | 13 | 5 | 3.4 | 44 |
027 | 20 | 7 | 23 | 5 | 2.9 | 66.5 |
aSP, spore count.
Clindamycin dosing achieved and maintained published biliary/faecal concentrations in the gut model during the instillation period. Germination of C. difficile PCR ribotype 027 occurred after clindamycin had decreased to subinhibitory concentrations (MIC = 2 mg/L). Faecal flora bacterial counts were unaffected by the addition of C. difficile PCR ribotype 027 to the gut model. Decreases in counts of facultative anaerobes, total anaerobes, Bacteroides spp., bifidobacteria, lactobacilli and clostridia were observed during the clindamycin instillation period; bifidobacteria and bacteroides counts decreased most markedly (data not shown). Counts of all faecal bacteria returned or slightly exceeded their pre-clindamycin exposure levels, excepting bifidobacteria, which remained undetectable following clindamycin instillation in the C. difficile PCR ribotype 027 experiment. Numbers of enterococci increased in both C. difficile PCR ribotype 001 and 027 experiments, from 4 to 10 log10 units following cessation of clindamycin dosing, and then remained high.
Effects of metronidazole on C. difficile PCR ribotypes 001 and 027 following induction of toxin production by clindamycin
Clindamycin administration in the CDI gut model caused germination and toxin production by both C. difficile 001 and 027 and gut flora changes as observed in the above experiments, and described previously.25 Metronidazole administration was commenced after 2 consecutive days of high cytotoxin titres (≥5 RU) in the gut model. C. difficile cytotoxin production was not observed in vessel 1 of either the C. difficile PCR ribotype 001 or 027 experiments thereafter. C. difficile 001 and 027 TVC decreased to spore levels in vessel 2 of each model, with concomitant declines in cytotoxin titres, when the metronidazole concentration was at least equivalent to the MIC for the strains (0.5 mg/L). Once the metronidazole concentration became subinhibitory, a further cycle of spore germination, proliferation and cytotoxin production occurred in vessel 2 of each model (Figure 2a; results not shown for C. difficile PCR ribotype 001). Cytotoxin titres declined in vessel 3 of the C. difficile 001 model, but not in the 027 experiment during this period (Figure 2b and c). Despite only a modest decrease in numbers of C. difficile PCR ribotype 001 in vessel 3, toxin titres decreased while the metronidazole concentration was at least the MIC (Figure 2b). Once the metronidazole concentration became subinhibitory for C. difficile PCR ribotype 001, germination and toxin production were again observed (Figure 2b). By contrast, there was no decrease in cytotoxin titres in vessel 3 of the C. difficile PCR ribotype 027 gut model experiment; these were maintained at a high level until the end of the experiment (Figure 2c). The median proportions of the TVC during the first 21 days of the germination periods of each strain that were accounted for by vegetative cells were 25 and 48% for C. difficile PCR ribotypes 001 and 027, respectively (P ≤ 0.001 Wilcoxon signed ranks test).
Each model was instilled with metronidazole in order to achieve a concentration equivalent to published faecal levels (9.3 g/L).29 However, measured metronidazole concentrations were markedly lower than expected: the concentrations were 25–75%, 8–9% and 3–7% of expected levels in vessels 1, 2 and 3, respectively (Table 4). Following metronidazole addition after 2 days of high toxin titres (≥ 5 RU), faecal bacterial populations initially decreased sharply in numbers, followed by slow decline. In both gut models, there were modest declines in bifidobacterial and bacteroides populations during the metronidazole installation period, but these were re-established as metronidazole concentrations waned (results not shown).
. | Vessel 1 . | Vessel 2 . | Vessel 3 . | |||
---|---|---|---|---|---|---|
Day of instillation . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . |
3 | 2.77 | 4.23 | 0.00 | 0.32 | 0.00 | 0.32 |
4 | 4.64 | 6.98 | 0.68 | 0.80 | 0.00 | 0.32 |
5 | 2.69 | 5.39 | 1.00 | 1.32 | 0.60 | 0.32 |
6 | 2.51 | 7.25 | 1.43 | 0.80 | 0.62 | 0.32 |
7 | 2.35 | 7.10 | 0.71 | 1.00 | 0.64 | 0.35 |
. | Vessel 1 . | Vessel 2 . | Vessel 3 . | |||
---|---|---|---|---|---|---|
Day of instillation . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . |
3 | 2.77 | 4.23 | 0.00 | 0.32 | 0.00 | 0.32 |
4 | 4.64 | 6.98 | 0.68 | 0.80 | 0.00 | 0.32 |
5 | 2.69 | 5.39 | 1.00 | 1.32 | 0.60 | 0.32 |
6 | 2.51 | 7.25 | 1.43 | 0.80 | 0.62 | 0.32 |
7 | 2.35 | 7.10 | 0.71 | 1.00 | 0.64 | 0.35 |
. | Vessel 1 . | Vessel 2 . | Vessel 3 . | |||
---|---|---|---|---|---|---|
Day of instillation . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . |
3 | 2.77 | 4.23 | 0.00 | 0.32 | 0.00 | 0.32 |
4 | 4.64 | 6.98 | 0.68 | 0.80 | 0.00 | 0.32 |
5 | 2.69 | 5.39 | 1.00 | 1.32 | 0.60 | 0.32 |
6 | 2.51 | 7.25 | 1.43 | 0.80 | 0.62 | 0.32 |
7 | 2.35 | 7.10 | 0.71 | 1.00 | 0.64 | 0.35 |
. | Vessel 1 . | Vessel 2 . | Vessel 3 . | |||
---|---|---|---|---|---|---|
Day of instillation . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . | PCR ribotype 001 . | PCR ribotype 027 . |
3 | 2.77 | 4.23 | 0.00 | 0.32 | 0.00 | 0.32 |
4 | 4.64 | 6.98 | 0.68 | 0.80 | 0.00 | 0.32 |
5 | 2.69 | 5.39 | 1.00 | 1.32 | 0.60 | 0.32 |
6 | 2.51 | 7.25 | 1.43 | 0.80 | 0.62 | 0.32 |
7 | 2.35 | 7.10 | 0.71 | 1.00 | 0.64 | 0.35 |
Discussion
The emergence of C. difficile PCR ribotype 027 as an epidemic strain in N. America and Europe has provoked considerable speculation as to the nature of its apparent increased virulence. There is evidence that C. difficile PCR ribotype 027 may cause more severe disease, and is associated with outbreaks with higher case-fatality ratios. The strain may also be associated with reduced metronidazole efficacy.15,17,18 Crucially, however, there is a paucity of robust, well-controlled data to explain the aetiology of the increased microbial virulence and apparently poor therapeutic efficacy. The present study sought to address this knowledge gap using a previously described, reproducible and clinically relevant in vitro gut model of CDI.23–26 The gut model provides gut-reflective conditions, but avoids the ethical and practical issues associated with volunteer, animal and batch culture (test-tube) studies.31 We have previously demonstrated C. difficile germination and cytotoxin production in response to clindamycin,25 and cefotaxime (with and without its active metabolite, desacetylcefotaxime),24 antimicrobials with marked propensity to induce CDI. In contrast, germination and cytotoxin production were not observed in the gut model following dosing with piperacillin-tazobactam, an agent uncommonly associated with the disease.23 These studies indicate that gut antimicrobial concentrations play an important role in the precipitation of C. difficile germination and cytotoxin production, while the role of altered colonization resistance, secondary to gut flora perturbation, in causing CDI is less clear than previously believed.
In the present study, we demonstrated that C. difficile germination and then cytotoxin production occurred after clindamycin concentrations decreased to subinhibitory levels, and notably after the recovery of most faecal flora to pre-dosing counts. These observations highlight the direct role of inducing antibiotic on C. difficile toxin production, as opposed to an effect mediated primarily through reduced colonization resistance. Warny et al.20 reported increased toxin production by C. difficile PCR ribotype 027 in batch culture in comparison with other C. difficile strains, and attributed this to an 18 bp deletion within tcdC, which is believed to negatively regulate toxin production in C. difficile. The authors observed statistically significantly increased growth by C. difficile PCR ribotype 027, and did not exclude this phenomenon as a possible explanation of this phenomenon. These data should be interpreted with caution, due to the experimental and analytical approaches of that study.31C. difficile behaviour in batch culture is poorly reproducible, and these conditions are not reflective of the gut environment (J. Freeman, S. Baines, W. N. Fawley and M. H. Wilcox, unpublished data).32 In contrast, the gut model is a continuous culture system, which allows C. difficile to be examined sequentially in conditions that mimic the retention times and elution rates of the human gut.23–26 Our results part contradict and extend those of Warny et al.20 The clear difference that we observed between the two epidemic strains was the duration of toxin production by C. difficile PCR ribotype 027, as opposed to a difference in rate or peak titre. C. difficile PCR ribotype 027 remained in its vegetative form and produced cytotoxin for 10 days longer (23 versus 13 days) than the comparator strain.
The extended vegetative phase of growth by C. difficile PCR ribotype 027 suggests that this strain is more successful at germination, at least under these gut model conditions, than C. difficile PCR ribotype 001. This property may afford C. difficile PCR ribotype 027 an advantage in eliciting prolonged inflammation and so possibly release of nutrients. These data indicate that the normal growth cycle of C. difficile PCR ribotype 027 is disrupted in some way. The temporary decline and subsequent increase in toxin titres during the extended vegetative growth phase of C. difficile PCR ribotype 027 may also indicate a biphasic growth cycle of this strain. Earlier toxin production following germination of C. difficile PCR ribotype 027 (1 day versus 2 days post-germination) probably reflects the greater number of vegetative cells within the total C. difficile population rather than a shift in the onset of toxin production during the growth cycle, as has been suggested.20 The increased virulence of C. difficile PCR ribotype 027 may be due to a deletion and more specifically an aberrant stop codon in the putative negative regulator of toxin production (tcdC).15–20,33 Sequence analysis confirmed the presence of two deletions in the tcdC region of C. difficile ribotype 027 strain (but not in C. difficile PCR ribotype 001) (data not shown). An 18 bp deletion and a further single nucleotide deletion leading to a stop codon were identified, equivalent to those recently identified by Curry et al.33 truncating the predicted TcdC to 65 amino acid residues. It is plausible that disruption of gene expression results in extended duration toxin production, rather than increasing the rate or amount of toxin produced, and thus leads to severe inflammation of the gut mucosa.
Conventional antimicrobial therapy of CDI has changed little since the disease was first described. Metronidazole and vancomycin treatment of CDI have similar efficacy in terms of initial response and recurrence rates.21,34 Metronidazole is generally used as first-line therapy given its relative low cost and probable reduced risk of glycopeptide-resistant enterococci selection. Recent reports suggested reduced initial response and increased recurrence rates in CDI patients treated with metronidazole compared with early efficacy studies, possibly associated with C. difficile PCR ribotype 027,21,22, However, Pépin et al.35 subsequently established that metronidazole was not inferior to vancomycin in the treatment of a first recurrence of CDI. We have previously observed that vancomycin treatment of clindamycin-induced C. difficile germination and toxin production in the gut model caused an immediate decrease in C. difficile total counts leaving only spores.25 In the present study, the most notable inhibitory effects of metronidazole were seen in the proximal gut model (vessel 1). In the more distal sections of the gut model, these effects were progressively less evident, and correlated with the measured concentrations of metronidazole. Decreases in C. difficile populations were only observed during the time that the metronidazole concentration was at least equivalent to the MIC of the C. difficile strain. Conversely, following initial decreases in cytotoxin titres produced by both C. difficile strains in vessel 2, these increased once the metronidazole concentrations fell below MIC levels. This same pattern was seen with C. difficile PCR ribotype 001 in the distal section of the gut model. However, metronidazole administration was not associated with reduced cytotoxin titres achieved by C. difficile PCR ribotype 027 in this distal section. Cytotoxin titres remained high here until the end of the experiment. This phenomenon was likely due to the failure to achieve inhibitory concentrations of metronidazole in vessel 3 of the C. difficile PCR ribotype 027 gut model.
Metronidazole concentrations in both C. difficile gut models were surprisingly low, despite dosing calculated to achieve 9.3 mg/L (published faecal concentrations).29 The highest concentrations were observed in vessel 1, but represented only 27% (median days 3–7) and 75% (median days 2–7) of expected faecal levels in C. difficile PCR ribotype 001 and 027 gut models, respectively. Metronidazole concentrations in vessels 2 and 3 were even lower. These poor concentrations of metronidazole cannot be explained by loss of drug due to activity on anaerobic gut flora, as this was minimal in all gut model vessels. Notably, measured clindamycin concentrations were very close to predicted levels, and despite marked effects on gut flora. Metronidazole concentrations decreased from vessels 1 to 3, precluding the possibility of washout and suggesting loss of metronidazole in each vessel. It is possible that metronidazole may have been inactivated by gut flora. Reports have described metronidazole inactivation by human gut bacteria, in particular enterococci.36,37 Nagy and Földes36 reported inactivation of metronidazole by both whole cells and cell-free extracts of Enterococcus faecalis. Rafii et al.37 also described metabolism of metronidazole to inactive compounds by Enterococcus gallinarum and Enterococcus casseliflavus gut isolates. Interestingly, we found that numbers of enterococci in the gut model increased (∼5 log10 cfu/mL) more than for any other bacterial group investigated during clindamycin instillation, and were maintained thereafter.
Inactivation of metronidazole by enterococci and/or other gut flora components could have important clinical implications in the context of the treatment of CDI. Inter-subject variation in gut flora and/or the spectrum of activity of the antibiotic(s) precipitating CDI, in terms of inhibition or selection of gut flora components, could plausibly influence the chance of symptomatic recurrence following metronidazole treatment. Resistance of C. difficile to metronidazole as a possible cause of treatment failures appears to be rare.38 While most symptomatic recurrences of CDI are due to re-infections,34 some may be relapses caused by the original infecting strain. This possibility could arise as a result of subtherapeutic metronidazole concentrations in the gut lumen following inactivation by gut bacteria, particularly as diarrhoea resolves and trans-mucosal passage of antibiotic decreases.29 We note that the efficacy of high dosages of metronidazole has not been examined in CDI. These issues deserve further study. The gut model does not reflect in vivo immunological or secretory events. Such factors may influence response to antimicrobial treatment of CDI, although this has not been elucidated.
The apparent induction of germination and subsequent cytotoxin production by metronidazole is not surprising, in the light of the poor drug concentrations achieved. We have previously highlighted the potential association between the gut lumen concentration of an antibiotic and the susceptibility of C. difficile to the drug.23–26 It is possible that the gut concentration of an antimicrobial relative to the MIC for C. difficile is a key factor in determining if germination and cytotoxin production occurs. If the antibiotic concentration is, or decreases to subinhibitory levels before the pathogen is effectively eliminated, then induction of toxin production may occur. We have demonstrated here and in previous studies that clindamycin causes marked depletion of the gut flora, but these recover to original numbers before C. difficile germination and toxin production are observed.25 The precise role of colonization resistance in the pathogenesis of CDI therefore remains unclear.
In summary, clear phenotypic differences exist between epidemic C. difficile PCR ribotype 001 and ribotype 027 strains. C. difficile PCR ribotype 027 germinates more readily, remains in a vegetative form and produces cytotoxin for substantially longer than C. difficile PCR ribotype 001 in the human gut model. These results may explain the increased severity of symptoms and higher case-fatality ratio associated with infection due to this emerging C. difficile strain. We demonstrated poor efficacy of metronidazole in the gut model. These results provide further evidence that gut antimicrobial concentrations, and the relative susceptibility of the infecting C. difficile strain, may be of more importance than depletion of colonization resistance in the pathogenesis of CDI. Unexpected poor concentrations of metronidazole, possibly due to inactivation by faecal bacteria, suggest that higher dosages of this drug should be explored to reduce the chance of therapeutic failures.
Acknowledgements
C. difficile PCR ribotype 027 strains from the US and UK were supplied by Dr J. S. Brazier, with the kind permissions of Dr Rob Owens (Maine Medical Center, Portland, USA) and Dr Jean O'Driscol (Stoke Mandeville).
Transparency declarations
M. H. W. has received honoraria for consultancy work, financial support to attend meetings and research funding from Astra- Zeneca, Bayer, Genzyme, Nabriva, Pfizer, Vicuron and Wyeth.