Subpopulations of Helicobacter pylori Are Responsible for Discrepancies in the Outcome of Nitroimidazole Susceptibility Testing

Abstract

Metronidazole susceptibility testing by E test was compared to that by disk diffusion for 263 Helicobacter pylori isolates and to that by breakpoint agar dilution for 90 H. pylori isolates. In 5% and 6% of the cases, respectively, results were discrepant. For each of 52 clinical isolates an E test was performed on 10 separate colonies. Subpopulations of resistant and susceptible bacteria were found in five cases. From three isolates, each colony was subcultured and tested up to 10 times. All but 1 of 292 tests showed the same result. We conclude that the E test is reliable and that subpopulations are responsible for discordant results.

Nitroimidazoles are used to treat Helicobacter pylori infection (22). The relevance of in vitro nitroimidazole resistance (NIR) of H. pylori to treatment efficacy is still debated (2, 5, 17, 20, 24). Furthermore, data on the evolution of NIR prevalence are conflicting (6, 14–16, 23). As nitroimidazole susceptibility testing is not standardized, methodology may be at least partly responsible for these controversies (17). In addition, susceptibility testing may also be influenced by the fact that patients can be infected with both susceptible and resistant bacteria (25). In this study the relevance of this phenomenon to E-test results was evaluated.

Antral biopsy specimens were rubbed on two types of selective agar plates, one containing Belo-horizonte medium (BHM medium) and the other containing campylobacter-selective medium. The plates were incubated at 36°C under microaerophilic conditions and were examined after 2 to 3 days. Colonies were confirmed to be H. pylori by Gram staining and determination of enzyme activities (catalase, oxidase, and urease). For susceptibility testing, Columbia agar plates, supplemented with 7% horse blood but without antibiotics (H. pylori susceptibility testing medium [HPS medium]), were inoculated with a suspension of multiple colonies. Metronidazole susceptibility was tested by E test, disk diffusion method, and breakpoint agar dilution method. The plates were read after 2 to 3 days. For the E test (AB Biodisk, Solna, Sweden) strains were considered resistant if the MIC of metronidazole was above 8 μg per ml. For disk diffusion a 5-μg metronidazole disk (Mast Laboratories, Liverpool, United Kingdom) was used. Strains with an inhibition zone of less than 10 mm were regarded as resistant (19). For breakpoint agar dilution two types of plates, one containing HPS medium supplemented with metronidazole (8 μg per ml) and the other containing HPS medium without antibiotics as a control, were used. If H. pylori grew on both plates, the strain was considered resistant.

In the first part of the study, the E-test results were compared with those of disk diffusion for 263 consecutive H. pylori isolates and with those of breakpoint agar dilution in 90 other isolates. All isolates were obtained from different patients. In the second part, 52 isolates, each obtained from a single antral biopsy specimen from 1 of 52 patients, were investigated. From each isolate 10 to 11 separate colonies were subcultured, and an E test was performed on each subculture. Results were compared with the E-test results obtained by using multiple colonies. For 3 of these 52 isolates (2 with MIC of ≥256 μg per ml and 1 with MIC of 0.5 μg per ml as determined by E test with multiple colonies), the 10 to 11 colonies were also subcultured on BHM medium. Bacteria were harvested after 2 to 3 days and resuspended on HPS medium for the E test or on BHM medium for further subculture. After 2 to 3 days, bacteria were again harvested from the BHM medium and the same procedure was repeated. In this way, 32 colonies were each tested up to 10 times.

When the E-test results were compared to the disk diffusion results and breakpoint agar dilution results, discordance was found in 5% (95% confidence interval [CI], 3 to 9%) and 6% (95% CI, 2 to 12%) of cases, respectively: for the comparison between E test and disk diffusion, 166 strains were susceptible by both tests, 83 strains were resistant by both tests, 8 strains were resistant by E test but susceptible by disk diffusion, and 6 strains were susceptible by E test but resistant by disk diffusion; for the comparison between E test and breakpoint agar dilution, 60 strains were susceptible by both tests, 25 strains were resistant by both tests, 1 strain was resistant by E test but susceptible by agar dilution, and 4 strains were susceptible by E test but resistant by agar dilution. In most of the cases of discordance the differences between the E-test and disk diffusion results were not around the cutoff: bacteria were completely resistant by one test and fully susceptible by the other test (Table 1). When 10 to 11 different colonies of an isolate were studied, 34 biopsy specimens were found to harbor only susceptible bacteria and in 13 specimens only resistant bacteria were found. In all of these cases the results were congruent with the outcome of routine susceptibility testing performed by using multiple colonies. In five biopsy specimens (10%; 95% CI, 3 to 21%) both resistant and susceptible bacteria were found. When multiple colonies were used, three of these five isolates were found to be resistant and the other two were found to be susceptible (Table 2). In our last study, in which the E test was repeated up to 10 times for each of 32 single colonies obtained from three isolates, variations in MIC were frequently seen. For one strain, for example, the MIC varied between 0.032 and 2 μg per ml. However, only one strain that was initially designated as resistant (for which the MIC varied between 24 and ≥256 μg per ml) was reclassified as susceptible (MIC was 8 μg per ml) (1 of 292 tests, 0.34%; 95% CI, 0.01 to 1.84%).

TABLE 1.

MIC and inhibition zones of the 14 strains with discordant results in the study comparing E test to disk diffusion

Strain no.	E-test MIC (μg/ml)	Disk diffusion inhibition zone (mm)
1	0.125	No zone
2	0.25	No zone
3	0.25	No zone
4	4	No zone
5	5	No zone
6	8	No zone
7	24	16
8	>32	12
9	>32	14
10	>32	16
11	>32	17
12	>32	22
13	>32	26
14	>32	26

TABLE 2.

Results of the E test performed on 10 to 11 separate colonies as compared to the E test performed on multiple colonies for the five biopsy specimens containing both resistant bacteria and susceptible bacteria

Specimen no.	MIC (μg per ml)a
	Multiple colonies	Single colony no.
		1	2	3	4	5	6	7	8	9	10	11
1	2	≥256	0.38	1	1	1	≥256	1	0.75	0.75	1
2	≥256	0.19	0.38	0.25	≥256	0.38	0.5	0.75	—	≥256	—
3	≥256	≥256	0.19	≥256	32	≥256	≥256	1.0	0.25	≥256	32
4	8	192	8	≥256	—	128	1	0.5	1.5	1.5	16
5	≥256	≥256	0.75	1.5	48	≥256	0.5	≥256	≥256	0.5	1.5	1.5

^a—, denotes a missing test result due to strain death, very slow growth, or fungal overgrowth. 

It has generally come to be recognized that, in comparison with testing of H. pylori susceptibility to other antibiotics, testing of nitroimidazole susceptibility is problematic (3, 9, 10, 17, 21). In our study we found disagreement between the tests in approximately 5% of cases, as did Midolo et al. (18). Although differences in methodology may cause variations in test results (17), we think that this does not account for the discrepancies found in our study, namely, that isolates were classified as highly resistant by one test and fully susceptible by another test. As the reproducibility of the E test for testing of the offspring of a single bacterium was almost perfect, the most likely explanation for the discordant results is a coinfection with both resistant bacteria and susceptible bacteria.

In our population such a coinfection could be demonstrated in approximately 10% of the cases, when only one biopsy specimen was examined. This is not unexpected, as it is known from DNA fingerprinting that patients can be infected with two or more different strains (8, 11, 13). Moreover, even within the same strain different patterns of susceptibility to metronidazole have been found (7, 13). As a study in gnotobiotic piglets has shown a microclonal mode of growth with limited migration of bacteria between different sites of the stomach (1), it is likely that as the number of gastric biopsy specimens examined increases the greater will be the number of different subpopulations encountered (13).

It is remarkable that coinfections with both resistant bacteria and susceptible bacteria have not been described for other antibiotics (7). The special position of nitroimidazoles in this respect may be explained by the genetic basis of NIR (9). NIR is related to null mutations in the rdxA gene, which encodes a nitroreductase. This enzyme is not essential for bacterial survival, as no decrease in metabolic or growth capacity is observed in the absence of functional enzyme (12). When a nitroimidazole-containing therapy fails to eradicate H. pylori, resistant mutants will become the major population in the stomach. After the antibiotic pressure has been removed, the wild type may reappear, but it may have no significant survival advantage. This results in a stable coinfection with both resistant bacteria and susceptible bacteria. In contrast, clarithromycin resistance is associated with a point mutation in rRNA genes (4). Inefficient protein synthesis, due to the mutation, is likely to diminish the chances of survival in the absence of clarithromycin. Therefore, if clarithromycin-susceptible bacteria are present they will eventually constitute the majority of the population.

In conclusion, our study shows that the E test is reliable for nitroimidazole susceptibility testing of H. pylori. A coinfection with both resistant bacteria and susceptible bacteria, however, is not uncommon and may lead to discordance between the results of different tests.