Metronidazole resistance in Helicobacter pylori is due to null mutations in a gene (rdxA) that encodes an oxygen-insensitive NADPH nitroreductase


  • Avery Goodwin,

    1. Department of Microbiology and Immunology, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada.,
    Search for more papers by this author
  • Dangeruta Kersulyte,

    1. Departments of Molecular Microbiology and Genetics Washington University Medical School, 4566 Scott Avenue, St Louis, MO 63110, USA., Divisions of,
    Search for more papers by this author
  • Gary Sisson,

    1. Department of Microbiology and Immunology, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada.,
    Search for more papers by this author
  • Sander J. O. Veldhuyzen van Zanten,

    1. Gastroenterology,
    Search for more papers by this author
  • Douglas E. Berg,

    1. Departments of Molecular Microbiology and Genetics Washington University Medical School, 4566 Scott Avenue, St Louis, MO 63110, USA., Divisions of,
    Search for more papers by this author
  • Paul S. Hoffman

    1. Department of Microbiology and Immunology, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada.,
    2. Infectious Diseases, Department of Medicine, Faculty of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada, B3H 4H7.
    Search for more papers by this author

Paul S. Hoffman. E-mail; Tel. (902) 494 3889; Fax (902) 494 5125.


Metronidazole (Mtz) is a critical component of combination therapies that are used against Helicobacter pylori, the major cause of peptic ulcer disease. Many H. pylori strains are Mtz resistant (MtzR), however, and here we show that MtzR results from loss of oxygen-insensitive NADPH nitroreductase activity. The underlying gene (called ‘rdxA’) was identified in several steps: transformation of Mtz-susceptible (MtzS) H. pylori with cosmids from a MtzR strain, subcloning, polymerase chain reaction (PCR) and DNA sequencing. We also found that (i) E. coli (normally MtzR) was rendered MtzS by a functional H. pylori rdxA gene; (ii) introduction of rdxA on a shuttle vector plasmid into formerly MtzRH. pylori rendered it MtzS; and (iii) replacement of rdxA in MtzSH. pylori with an rdxA::camR null insertion allele resulted in a MtzR phenotype. The 630 bp rdxA genes of five pairs of H. pylori isolates from infections that were mixed (MtzR/MtzS), but uniform in overall genotype, were sequenced. In each case, the paired rdxA genes differed from one another by one to three base substitutions. Typical rdxA genes from unrelated isolates differ by ≈ 5% in DNA sequence. Therefore, the near identity of rdxA genes from paired MtzR and MtzS isolates implicates de novo mutation, rather than horizontal gene transfer in the development of MtzR. Horizontal gene transfer could readily be demonstrated under laboratory conditions with mutant rdxA alleles. RdxA is a homologue of the classical nitroreductases (CNRs) of the enteric bacteria, but differs in cysteine content (6 vs. 1 or 2 in CNRs) and isoelectric point (pI = 7.99 vs. 5.4–5.6), which might account for its reduction of low redox drugs such as Mtz. We suggest that many rdxA (MtzR) mutations may have been selected by prior use of Mtz against other infections. H. pylori itself is an early risk factor for gastric cancer; the possibility that its carcinogenic effects are exacerbated by Mtz use, which is frequent in many societies, or the reduction of nitroaromatic compounds to toxic, mutagenic and carcinogenic products, may be of significant concern in public health.


Metronidazole (Mtz) [1-(2-hydroxyethyl)-2-methyl-5-nitroimidazole] is a key component of combination therapies that are widely used against Helicobacter pylori (Malfertheiner et al., 1997), a microaerophilic, Gram-negative pathogen that is highly specific for the human gastric mucosa. H. pylori tends to establish chronic and often life-long infections that constitute a major cause of peptic ulcer disease and an important risk factor for gastric cancer, one of the most common malignancies worldwide (Correa, 1996). Most residents of developing countries are infected with H. pylori (Taylor and Parsonnet, 1995); this situation is ascribed to poor sanitation, which results in frequent exposure to the pathogen. In the US and Western Europe, the prevalence of infection is generally lower, and is correlated with socioeconomic status and age: some half of older adults but less than one-tenth of young children in these industrialized societies are H. pylori-infected (Taylor and Parsonnet, 1995; Dunn et al., 1997). Mtz resistance (MtzR) is an important variable in the treatment of H. pylori infections and its presence markedly reduces the efficiency of Mtz containing regimens (Chiba et al., 1992; Graham et al., 1992). The incidence of MtzR also varies geographically, with half or more of H. pylori strains from developing countries and some 10–30% of strains from the US and Western Europe being MtzR (Dunn et al., 1997; Veldhuyzen van Zanten et al., 1997). The incidence of MtzR among H. pylori isolates generally parallels the level of Mtz usage in a particular society. Thus, it is parsimonious to imagine that many of the H. pylori strains currently resistant to Mtz reflect the frequent use of Mtz and related nitroimidazoles for treatment of anaerobic and protozoan infections (Grunberg and Titsworth, 1974; Hof and Sticht-Groh, 1984; Edwards, 1993), but in dosing regimens that generally do not eliminate MtzSH. pylori from an infected person. Any inhibition of H. pylori growth during such periods of Mtz therapy would enrich or select MtzR strains.

The basis for susceptibility of wild-type H. pylori to Mtz and the mechanisms of resistance have been of interest and concern since the early days of H. pylori research (see, for example, McNulty et al., 1985; Glupczynski and Burette, 1990). Well-studied model organisms such as Pseudomonas aeruginosa and Escherichia coli, which are aerobic or facultatively anaerobic, are MtzR, whereas many anaerobes and microaerophiles are susceptible to Mtz (MtzS). MtzR is relatively rare in anaerobes (Rasmussen et al., 1997), and therefore, one might imagine that the high incidence of MtzR in microaerophiles is due to a mechanism of action that differs from that found in anaerobes. The available evidence from studies of protozoan and anaerobic bacterial species had suggested that Mtz toxicity to H. pylori might depend on its reduction to the nitro anion radical and other compounds including hydroxylamine (Moreno et al., 1983; Lindmark and Muller, 1976; Kedderis et al., 1988). Hydroxylamine is particularly damaging to macromolecules such as DNA and proteins (Lindmark and Muller, 1976; Kedderis et al., 1988). Under aerobic or microaerobic conditions, molecular oxygen could convert reduced Mtz (nitro anion radical) back to the parent compound by a process termed ‘futile cycling’, which essentially generates superoxide anions instead of hydroxylamine (Smith and Edwards, 1995). Because futile cycling has not been demonstrated experimentally (Smith and Edwards, 1995), reductions involving two and four electron transfers that favour hydroxylamine formation, such as would occur with ferredoxins and flavodoxins as electron donors, seemed very plausible, despite a lack of experimental evidence for direct enzymatic reduction of Mtz by H. pylori. Given this background, several possible mechanisms for MtzR in H. pylori merited consideration: decreased Mtz uptake or active efflux; deficiency in Mtz activation or modification; target modification or loss; and increased DNA repair or oxygen scavenging capabilities (Hoffman et al., 1996). Indeed, inactivation of recA, a gene needed for generalized DNA repair and recombination, greatly enhanced Mtz susceptibility of wild-type H. pylori (Thompson and Blaser, 1995) and a cloned recA gene from a MtzR strain seemed to increase the already very high level of resistance that E. coli exhibits (Chang et al., 1997).

Our initial investigations of metronidazole resistance focused on the metabolic enzymes of H. pylori ; in particular, on pyruvate:ferredoxin/flavodoxin oxidoreductase (POR) and α-ketoglutarate oxidoreductase (KOR) (Hoffman et al., 1996), in part because studies in anaerobes had shown POR to be responsible for Mtz activation (Moreno et al., 1983; Narikawa, 1986; Lockerby et al., 1985). Our studies showed that POR and KOR activities of MtzR strains of H. pylori were repressed in bacteria that had been cultured in the presence, but not in the absence, of Mtz (Hoffman et al., 1996). This result indicated that these reductases were regulated by Mtz, which is consistent with a model in which H. pylori POR and KOR mediate Mtz toxicity. However, those experiments did not test whether this, or any of several other changes that have been identified to date (see Hoffman et al., 1996; Smith and Edwards, 1997), is a primary effect, and the cause of resistance, or a secondary consequence of other metabolic perturbations that Mtz elicits. Similarly, although MtzR mutants are easily derived from many MtzS strains in the laboratory, the genetic basis for naturally occurring resistance, whether by mutation in a normal chromosomal gene or by acquisition of a new ‘resistance’ gene, was unknown.

Here we show that mutational inactivation of an H. pylori gene, which encodes an oxygen-insensitive NADPH nitroreductase, to be called rdxA (designated HP0954 in the entire genome sequence) (Tomb et al., 1997), is the cause of naturally acquired MtzR in H. pylori. We suggest that the MtzS phenotype of wild-type H. pylori is due to the reduction of Mtz by RdxA to hydroxylamine, which is directly responsible for killing of wild-type H. pylori. Expression of recombinant rdxA in E. coli, a species that is ordinarily resistant to > 300 μg ml−1 Mtz, rendered the bacteria MtzS. The inferred RdxA protein is a homologue of the classical nitroreductases (NfsB, NfnB and Cnr) of enteric bacteria (Watanabe et al., 1990; Bryant and DeLuca, 1991; Bryant et al., 1991; Zenno et al., 1996a), which is responsible for the reduction of nitroaromatic compounds, including Mtz, to toxic, mutagenic and carcinogenic products (Lindmark and Muller, 1976; Watanabe, et al., 1989; Zenno et al., 1994). The possibility of delivery of a potent mutagen to the gastric mucosa by H. pylori during any Mtz therapy has sobering implications for public health in light of findings that H. pylori infection can be a critical risk factor for gastric cancer.


Identification of a nitroreductase that confers Mtz sensitivity in H. pylori

The gene responsible for naturally occurring MtzR in H. pylori was sought using a strategy based on an earlier finding (Wang et al., 1993) that DNA from MtzR clinical isolates could transform MtzS strains to MtzR. To maximize the chance of finding the MtzR determinant, independent of whether naturally occurring MtzR is caused by a particular type of allele of a normal chromosomal gene, or by an added gene that is absent from the genomes of MtzS strains, we used a cosmid cloning approach. A cosmid library was constructed in the Lorist6 vector using a partial Sau3A digest of genomic DNA from MtzR strain 439, and E. coli transductants carrying the cosmids were arrayed in microtitre plates. Cosmid DNAs were prepared in batch from each microtitre plate and tested as 96-member pools for the ability to transform MtzS strain 500 to MtzR. Low-level transforming activity was found reproducibly in one of the nine pools tested (11 transformant colonies, vs. 15 in a control using 15 μg of strain 439 genomic DNA); no transformants were obtained with cosmid DNAs from any of eight other microtitre plates. The cosmid responsible for MtzR-transforming activity was identified in two more transformation steps: first using 12-member pools prepared from each of the eight rows in this microtitre plate; and then using individual cosmids from the one active row. EcoRI digestion generated four DNA fragments from the cosmid containing the MtzR determinant, and the MtzR-transforming activity was localized to one of them, a 2.3 kbp fragment. Two open reading frames (ORFs) were found in this cloned fragment by DNA sequencing. One had strong protein level homology to the gene for prolipoprotein diacyglycerol transferase lgt (corresponds to HP0955 in the entire H. pylori genome sequence) (Tomb et al., 1997) and seemed unlikely to be involved in MtzR. The other ORF had protein level homology to classical oxygen-insensitive NAD(P)H nitroreductases (CNRs) of several other Gram-negative bacteria (see Table 1) and was a good candidate because some of its homologues are known to reduce metronidazole or related compounds (Lindmark and Muller, 1976; Yamada et al., 1997). This H. pylori gene corresponds to the ORF designated HP0954 in the full genome sequence (Tomb et al., 1997) and, interestingly, exhibits 54% similarity with another ORF (HP0642) that encodes a NAD(P)H flavin nitroreductase (we suggest be named ‘frxA’) another CNR homologue. The inferred RdxA product from MtzR strain 439 is 196 amino acids long. PCR amplification and sequencing of the corresponding segment from the MtzS strain 500 revealed an ORF that is 14 codons longer at the 3′ end (210 codons, see Fig. 1 for rdxA). The rdxA gene from a MtzR transformant of strain 500 (strain HP1107) that was made with genomic DNA from strain 439 was identical in DNA sequence to that of the 439 parent strain (Fig. 2). These results indicate that MtzR in H. pylori can result from inactivation of rdxA, which in strain 439 occurred by a nonsense mutation that resulted in a truncated RdxA protein. The WT rdxA gene was 630 bp in length and had a Shine–Dalgarno sequence 5 bp upstream of the start codon. Although the CNR proteins of the enteric bacteria are acidic proteins, including HP0642 (pI = 5.4–5.6), and generally contain one to two cysteine residues, RdxA is a basic protein (pI = 7.99) and contains six cysteine residues. One of the cysteine residues (position 87) is conserved in the CNR proteins of the enterics, whereas the cysteine located at position 159 is in a motif (L/IDSCI/PI) shared with the inferred product of frxA (HP0642). Another motif common to all of the CNRs is QPWHF (PW absolutely conserved) located within a highly conserved region between positions 43–58 in RdxA.

Table 1. . Similarity of RdxA to other classical nitroreductases.Thumbnail image of
Figure 1.

. Nucleotide sequence and deduced amino acid sequence of rdxA of WT strain 500. The Shine–Dalgarno (SD) ribosome-binding site is underlined. The underlined amino acid sequence defines a highly conserved region among CNR proteins. Cysteine residues are highlighted in bold face and the SphI sites used for insertion of the camR cassette are underlined and noted. **H. pylori strains 439 and 1107 contain transition substitutions (TT for CC).

Figure 2.

. Location of amino acid substitutions in RdxA from matched MtzR/S strains and from clinical isolates. H. pylori strain 1107 was created by transforming DNA from MtzR strain 439 into MtzS strain 500. Note that the RdxA amino acid sequence is identical, indicating allelic exchange recombination occurred outside the rdxA locus. Other clinical isolates are included for comparison. The five matched pairs of isolates are grouped separately and the amino acid substitutions are listed in Table 3.

Nitroreductase activity and rdxA expression in E. coli

Cell-free extracts from MtzS and MtzR strains of H. pylori were screened for nitroreductase activity using standard assays that use either menadione or nitrofurazone as electron acceptors (Bryant and Deluca, 1991; Zenno et al., 1994) (data not presented). No significant differences in the nitroreductase activities of either isogenic pairs of MtzS and MtzR strains or of various clinical isolates were detected, suggesting that H. pylori most probably possesses multiple nitroreductases, which is supported by the full genome sequence (e.g. frxA, HP0642, Tomb et al., 1997), and by the precedent of multiple nitroreductases found in enteric bacteria (Zenno et al., 1996a,b). In addition, no Mtz reductase activity was detected in crude extracts from MtzS strains of H. pylori, independent of whether NADPH or NADH were used as electron donors, which is consistent with earlier observations (Hoffman et al., 1996). The inability to detect Mtz reductase activity in cell-free extracts of H. pylori might be attributable to oxidation of key components during the preparation or to an inability of the assays used to detect very low levels of Mtz reductase activity.

Because E. coli strains are intrinsically resistant to Mtz (> 300 μg ml−1), we explored the possibility that expression of rdxA in E. coli might render the organism susceptible to Mtz. We found that the cloned rdxA genes (rdxA cloned in a pBluescript vector, downstream of the lac promoter) from each of 8 MtzSH. pylori strains tested indeed rendered E. coli MtzS (killing by 10–60 μg ml−1) during aerobic growth on LB agar. In contrast, equivalent plasmid clones made with rdxA genes from each of eight MtzRH. pylori had no effect on the intrinsic high level of Mtz resistance of the E. coli host. Each of the strains used in the rdxA sequence analyses (Fig. 2) was tested in this way, yielding results that completely supported the use of in vivo assays in E. coli as a surrogate for monitoring the rdxA activity of H. pylori. In vivo assay of frxA (cloned from the 26695 strain of H. pylori into pBluescript) in E. coli indicated that the FrxA (flavin reductase) activity did not alter the intrinsic resistance of E. coli to Mtz.

The cloned rdxA gene from the H. pylori strain that rendered E. coli most susceptible to Mtz (strain 950) was tested for nitroreductase activity by spectrophotometric assay. Cell-free extracts from E. coli harbouring rdxA from this strain exhibited 40-fold higher than background NADPH-dependent nitroreductase activity using metronidazole as electron acceptor, and assayed by following either Mtz reduction or oxidation of NADPH (Table 2). No reductase activity was found using NADH instead of NADPH as the electron donor, nor was any detected using extracts of E. coli carrying pBluescript by itself or with an rdxA mutant (MtzR allele from strain 1043). These results indicate that RdxA protein can reduce Mtz and differs from other CNRs in showing specificity for NADPH. Among the known nitroreductases, only NfsA of E. coli shows specificity for NADPH (Zenno et al., 1996b), but this gene exhibits no DNA- or protein-level homology with RdxA (or with FrxA, HP0642) of H. pylori. Even in E. coli, Mtz reductase activity varied among clones and endogenous nitroreductase activity of E. coli interfered with RdxA nitroreductase assays, suggesting that further analysis of enzymatic properties of RdxA would benefit greatly from its purification. Nevertheless, our results indicate that expression of WT rdxA, but not frxA in E. coli, causes a marked increase in susceptibility to Mtz and support the conclusion that rdxA function is responsible for the MtzS of wild-type H. pylori, and that MtzR in this pathogen results from rdxA inactivation.

Table 2. . Metronidazole reductase activity of RdxA nitroreductase. Metronidazole reduction was measured in crude extracts of E. coli strain JF626 grown aerobically in LB broth. pBSK is pBluescript vector control; pBS950 is WT rdxA cloned into pBSK and pBS1043 is MtzR rdxA cloned in pBSK. The assay contained NADPH and Mtz. The enzymatic reaction was followed at 320 nm to measure Mtz reduction and at 340 nm to measure NADPH oxidation. The values are corrected for NADPH oxidase activity. No activity was found when NADH was used as substrate.Thumbnail image of

Sequence analysis of rdxA in closely related pairs of MtzR and MtzS strains

To assess how often MtzR is acquired by de novo mutation vs. rdxA gene transfer from an unrelated strain that is already MtzR, we studied the rdxA genes from infections that were mixed with respect to MtzR/MtzS, and in which the MtzR and MtzS isolates seemed to be very closely related based on arbitrarily primed PCR (Y. Valdez, V. Hernandez, D.E. Berg, K. Preiter, D.K. Bhasin, A. Ramirez-Ramos, S. Recavaren, R. Leon Barua and R.H. Gilman, in preparation; D. Kersulyte, H. Chalkauskas and D.E. Berg, in preparation). With each of five such pairs of isolates studied, the PCR amplified rdxA-containing segment obtained was about the same size (≈937 bp). This implied that resistance was due to point mutations and not to insertion, deletion or other rearrangement. DNA sequence analysis showed that the rdxA genes from MtzR and MtzS members of each pair were closely related but differed by 1–3 bp in the 630-bp-long gene (resulting in one or two amino acid replacements) in each case (see Fig. 2 and Table 3[link]). Because unrelated rdxA genes differed on average by about 5% (28–34 bp of 630 bp), this indicates that MtzR was due to de novo mutation, not horizontal gene transfer from another strain.

Table 3. . Types of point mutations in matched pairs of MtzR and MtzS strains and amino acid substitutions. a. Comparison of divergence in rdxA of unrelated H. pylori strains 439 and 500. Listed are the number of amino acid changes between these strains.Thumbnail image of

Four of the five alleles resulted in single amino acid changes in the inferred 210-amino-acid-long RdxA protein: G→V at position 145 in mutant 10amt3; A→T at position 180 in 12mtz; R→G at position 200 in H2mt; and K→E at position 63 in strain 21cmt. The fifth rdxA mutant allele (B1amt) would encode a protein with two amino acid sequence changes, Y→C at position 47, which is in a region that is highly conserved in CNRs (position 43–57), and also A→T at position 143.

rdxA-inactivation is sufficient for MtzR: allelic exchange mutagenesis and complementation

Based on finding non-functional rdxA alleles in each MtzR clinical isolate studied, we tested whether rdxA inactivation is also sufficient for resistance, or whether additional mutations are also needed. A CmR cassette was inserted into the SphI site in a cloned rdxA gene from MtzS strain H2csr, and this mutant allele was transformed via a non-replicating vector (pBluescript) into MtzS strain 26695, with selection for the CmR marker. Each of the 30 CmR transformants tested was able to grow on Mtz-containing medium (18 μg ml−1 Mtz), and thus had acquired high-level MtzR. This showed that simple inactivation of rdxA is sufficient for MtzR in H. pylori.

Previous studies had shown that growth of MtzR strains in Mtz-containing medium resulted in disappearance of POR activity, another enzyme that putatively can reduce Mtz, and therefore that should render H. pylori MtzS whenever it is active (Hoffman et al., 1996). In the present experiments, we found that growth of the rdxA::camR insertion mutant strain (which had been selected solely by its CmR phenotype) in Mtz-containing medium also resulted in the disappearance of POR activity. In addition, during growth in Mtz-free medium, this mutant strain exhibited only half as much POR activity as its isogenic rdxA+ (MtzS) parental strain. These studies suggest that mutations in rdxA may indirectly affect the level of POR activity through an as yet poorly understood but potentially important mechanism.

In complementary experiments, the rdxA gene from the MtzS strain 500 was PCR amplified and cloned into pDH26, a CmR shuttle vector that is stably maintained in H. pylori (obtained from R. Haas), and the construct was transformed into the MtzR strain 1061R. Strain 1061 had been made MtzR by transformation of mutant rdxA allele originating from MtzR strain 439. Each of the eight CmR colonies tested exhibited a MtzS phenotype, and rdxA-containing plasmid DNAs were easily reisolated from each of them; this indicates that the rdxA nonsense mutant allele is recessive, as expected. These results further establish that null mutations in just a single gene, rdxA, are responsible for MtzR in H. pylori.


We have examined the basis of susceptibility and resistance to the antimicrobial agent metronidazole (Mtz) in H. pylori. Our experiments indicate (i) that the toxicity of Mtz for H. pylori depends on its reduction to hydroxylamine by an oxygen-insensitive, chromosomally encoded NADPH nitroreductase (rdxA ; HP0954 in the genome database) (Tomb et al., 1997); (ii) that resistance results from mutational inactivation of rdxA and not from the acquisition of foreign resistance genes (in contrast to common mechanisms of resistance against other antibiotics and bacterial species) (Levy, 1992); (iii) that newly arisen MtzR derivatives of a MtzS strain generally arise by de novo mutation, rather than from the acquisition of mutant rdxA alleles from other already MtzRH. pylori strains; and (iv) that RdxA activity is dispensable. In accord with this, MtzR strains display no significant changes in metabolic or growth capacity compared with isogenic MtzS strains in culture (Hoffman et al., 1996). Further tests are needed to learn whether loss of RdxA function has any significant effect on virulence.

Four results established the importance of a functional rdxA gene in MtzS, and rdxA inactivation as the general mechanism of MtzR in H. pylori. First, a mutant allele of rdxA was found using a DNA transformation strategy: one cosmid in a library made from a MtzR clinical isolate was found to transform a MtzS recipient to MtzR; subcloning from this cosmid, and further transformation identified the segment responsible for MtzR, and DNA sequencing revealed rdxA, a nitroreductase gene with significant protein level homology to the CNRs of enteric bacteria. The allele of rdxA that was responsible for transformation of the MtzS strain to MtzR in these first experiments contained a nonsense (translational stop) codon 14 codons before the 3′ end of the ORF (as defined by sequences of rdxA genes from MtzS strains). Second, E. coli, which is normally MtzR, was rendered MtzS by cloned rdxA genes from each of 8 MtzSH. pylori strains, but not by cloned rdxA genes from any of 8 MtzRH. pylori strains that were tested. This showed that MtzR strains contain mutant (inactive) rdxA genes. DNA sequencing showed that point mutations (missense and nonsense) at other sites in rdxA were responsible for rdxA inactivation in these strains. Third, introduction of rdxA from a MtzSH. pylori strain on a shuttle vector plasmid rendered a formerly MtzR recipient strain MtzS; this further illustrates that a functional RdxA nitroreductase contributes to the MtzS phenotype of normal H. pylori. Fourth, H. pylori derivatives with camR inserts in their rdxA genes, and that had been selected solely by their CmR phenotype, exhibited a typical MtzR phenotype. Collectively, these results showed that a functional RdxA nitroreductase is key to the normal MtzS phenotype of wild-type H. pylori, and, conversely, that rdxA inactivation is necessary and sufficient for MtzR in this species.

Nitroreductases from other organisms are classified as oxygen sensitive or insensitive based on whether the substrates are reduced in one- or two-electron transfer reactions respectively. One-electron transfer reductions of the nitro group of a particular compound produces the nitro-anion radical, which in the presence of oxygen generates superoxide anions and regeneration of the 5-nitro group (Moreno et al., 1983; Edwards, 1993). It has been suggested that aerobes and facultative anaerobes are resistant to Mtz because under aerobic conditions redox cycling leads to regeneration of Mtz (Smith and Edwards, 1995). Indeed, the MtzS of Actinobacillus actinomycetemcomitans under anaerobic, but not aerobic conditions, is consistent with the concept of redox cycling and the nitroreductase activity implicated in MtzS of this species may be of the oxygen-sensitive type (Pavicic et al., 1995). In contrast, the MtzS of H. pylori was not affected by growth under different oxygen tensions; this suggests that one electron transfer is probably not involved in Mtz reduction in this microaerophilic bacterium (Smith and Edwards, 1995), an interpretation supported by our finding that an oxygen-insensitive nitroreductase is responsible for the MtzS of H. pylori. Microaerophiles in general are susceptible to Mtz, and display patterns of resistance similar to those noted for H. pylori (Hoff and Stricht-Groh, 1984; Lariviere et al., 1986), suggesting that homologues of rdxA may be found in these other species.

The majority of nitroreductases thus far studied are of the oxygen-insensitive type and are capable of reducing nitroaromatic compounds through sequential two-electron reductions, resulting in nitroso intermediates and hydroxylamine end products (Lindmark and Muller, 1976; Bryant and Deluca, 1991). This interpretation is supported by the direct demonstration that the enteric homologues of RdxA (CNRs NfsB of E. coli, Cnr of Salmonella typhimurium, and NfsB of E. cloacae) reduce 4- and 5-nitro compounds by two-electron transfer reactions (Bryant and Deluca, 1991; Zenno et al., 1996a; Yamada et al., 1997). The substrate specificity of the CNRs is often a function of the redox potential of the 5-nitro group (Bryant and Deluca, 1991), and in this regard the intrinsic resistance of enteric bacteria to Mtz is due to the very low redox potential of Mtz (Narikawa, 1986). However, reduction of Mtz and other nitroaromatic compounds to mutagenic end products by S. typhimurium has been demonstrated in the Ames test (Lindmark and Muller, 1976). Null mutations in the S. typhimurium gene for Cnr, an rdxA homologue, renders S. typhimurium resistant to the mutagenic effects of nitro-containing compounds (Yamada et al., 1997). We suggest that CNR activates Mtz in these microbes, generating hydroxylamine at levels that are too low to cause much lethality yet are still sufficient for mutagenesis. The MtzS of E. coli strains containing cloned H. pylori rdxA genes, for which Mtz reductase activity was measured in two strains, suggests that lethality must be due to the greater production of hydroxylamine from Mtz by the H. pylori RdxA nitroreductase.

The H. pylori RdxA nitroreductase is distinctive from the enteric CNR homologues in both cysteine content (six vs. one or two in CNRs) (Bryant et al., 1991; Zenno et al., 1996a; Tomb et al., 1997) and in abundance of lysine and arginine, which contribute to an alkaline pI (7.99 vs. 5.4–5.6 for CNRs). Although the CNRs generally oxidize either NADH or NADPH in a flavin cofactor-dependent reaction, RdxA showed a preference for NADPH. Spectral analysis of RdxA in extracts of E. coli have not confirmed the presence of a flavin cofactor. Despite these differences, RdxA contained a conserved amino acid motif common to the CNRs (QPWHF) as well as the positioning of a strategic cysteine residue (position 87). We suggest that the multiple cysteine residues of RdxA together with the more alkaline nature of the protein may contribute to both a lower redox potential and a greater substrate specificity of this enzyme for Mtz. These properties might be achieved through the formation of disulphide bonds or the chelation of metal cofactors, which might form a flavin-independent catalytic centre. It has been suggested that a disulphide bond of the CNR homodimer may participate as an electron acceptor in the oxidation of NAD(P)H (Inouye, 1994; But see Zenno et al., 1996a) and in an alkyl hydroperoxide reductase from S. typhimurium, two cysteine residues participate in catalysis (Ellis and Poole, 1997). Further analysis of the potential role of disulphide bonds or flavin cofactors in RdxA activity will require purification of the protein.

Typically, bacteria contain several different nitroreductases including flavin and ferredoxin reductases that may exhibit nitroreductase activity (Zenno et al., 1996a,b). One relatively close homologue of rdxA, with 25% protein-level identity over 181 amino acids, is frxA (HP0642), which encodes a NAD(P)H flavin reductase (FrxA) similar to the flavin reductase of Haemophilus influenzae (Tomb et al., 1997). The results presented above suggest that FrxA does not contribute significantly to Mtz susceptibility or resistance; in support of this, we have also found that the frxA gene cloned in the pBluescript plasmid vector does not affect the intrinsic high resistance of E. coli to Mtz. As even MtzR strains of H. pylori become susceptible to Mtz under anaerobic conditions (Smith and Edwards, 1995), perhaps FrxA and/or other ferredoxin and flavin reductases, such as those found in the annotation of the H. pylori genome sequence (Tomb et al., 1997), may contribute to the activation of Mtz under anaerobic conditions.

Naturally occurring MtzR is associated with a Mtz-inducible depression of activity of pyruvate oxidoreductase (POR) and as little as 3–5 μg ml−1 of Mtz in the culture medium is sufficient to abolish POR activity of MtzR strains (Hoffman et al., 1996). This depression of POR was also seen in a MtzR strain containing a camR insertion in rdxA, a strain selected by its CmR, not by its MtzR. This result indicates that repression of POR activity is not due to a secondary mutation selected by enhancement of MtzR. Based on studies with anaerobes, POR should also be capable of reducing Mtz (Lockerby et al., 1985), and we propose that the POR of H. pylori acts similarly. This thinking suggests that POR activity could be responsible for the transient growth inhibition and limited killing seen when MtzRH. pylori are first exposed to Mtz (Lacey et al., 1993). The ability to turn off synthesis or accumulation of POR in response to Mtz might then be an important component of resistance. Just how this putative regulatory mechanism operates is not yet known, but it is attractive to imagine that it involves a response to the chemical (hydroxylamine induced) damage to DNA, protein or other macromolecules, analogous to the bacterial response to alkylation damage (see Volkert, 1988; 1989). Such a mechanism might also be advantageous during normal growth (without Mtz treatment), helping safeguard H. pylori against deleterious effects of reduction of other nitroaromatic compounds that might be encountered in situ such as hydroxylamine adducts that might result from the action of nitric oxide with amines.

In separate studies of H. pylori from human populations at high risk of infection (Peru, Lithuania), we had identified pairs of strains, one MtzR and one MtzS, that were closely matched in RAPD fingerprint (Berg et al., 1997; Y. Valdez, V. Hernandez, D.E. Berg, K. Preiter, D.K. Bhasin, A. Ramirez-Ramos, S. Recavaren, R. Leon Barua and R.H. Gilman, in preparation; D. Kersulyte, H. Chalkauskas and D.E. Berg, in preparation). Although rdxA genes from unrelated strains differed in DNA sequence by 5% on average, the rdxA genes from MtzS and MtzR isolates from the same person differed by only one or a few base substitutions. This result indicated that MtzR resulted from de novo mutation, and not by gene transfer from an unrelated MtzR strain, even though at least transiently mixed infection seems to be quite common in these high risk (Peruvian and Lithuanian) societies.

Mtz is widely used to treat many bacterial and protozoan infections in the Third World as well as in industrialized nations, generally in dosing regimes that should inhibit H. pylori growth and thereby select resistant mutants but that do not eradicate the organism. Given the formation of hydroxylamine, a potent mutagen, by the RdxA nitroreductase of H. pylori, we suggest that many of the mutations to MtzR were induced by the Mtz therapy (even when given for other infections) and that they are not of spontaneous origin. Accordingly, we view with concern the possibility that MtzSH. pylori might deliver mutagenic levels of hydroxylamine to gastric epithelial tissue whenever Mtz is used or nitroaromatic compounds are ingested, particularly in light of the mutational origins of human cancers, and the importance of H. pylori infection as an early risk factor for gastric cancer.

Experimental procedures

Bacterial strains and growth conditions

The H. pylori isolates used in this study were isolated from human gastric biopsy samples and were obtained from the Victoria General Hospital, Halifax, Nova Scotia, Canada, and have been previously described (Hoffman et al., 1996). Paired MtzR and MtzS strains from the same patient that we found to be closely matched in overall genotype had been isolated from biopsies from Peruvian and Lithuanian patients, which were kindly provided by Drs R. H. Gilman and H. Chalkauskas respectively. Bacterial strains were grown at 37°C on Brucella agar plates supplemented with 10% fetal calf serum (FCS) in a microaerobic incubator maintained at 7% O2, 5% CO2. Liquid cultures were grown in Brucella broth with 10% FCS in 125 ml screw-capped flasks; the medium was equilibrated with 7% O2, 5% CO2 in the microaerobic incubator for 1 h before inoculation, and then the flasks were sealed and placed on a rotary shaker at 150 r.p.m. Unless otherwise indicated, metronidazole-resistant strains were grown with 18 μg ml−1 of Mtz, which is one half the minimal inhibitory concentration. Bacteria were harvested by centrifugation after 3–4 days of growth, and either used immediately or stored as a pellet at −70°C. E. coli strains DH5α (BRL) and ER1793 (New England Biolabs) were grown on Luria–Bertani (LB) agar plates supplemented with the appropriate antibiotics.

Cosmid library construction and screening

Genomic DNA was prepared from MtzR strain 439 and partially digested with Sau3A to generate a population of DNA fragments in the 20–45 kb range, as described previously (Hoffman et al., 1989). These DNA fragments were cloned into BamHI-cleaved Lorist6, a cloning vector that has been useful for making cosmid libraries from other H. pylori strains, and the ligated DNA was packaged into λ phage particles (Bukanov and Berg, 1994). Cosmids were recovered after infection of E. coli ER1793, which is deficient in restriction/modification systems, and transductants carrying cosmid clones were selected on LB agar containing 30 μg ml−1 of kanamycin. KanR colonies were picked into wells in microtitre dishes. Cosmid DNAs were prepared in batches from the growth of 96 clones per microtitre plate in 50 ml of LB broth and the cosmid DNA was purified by miniprep (Sambrook et al., 1989). Nine batches of cosmid DNA were screened for those carrying a MtzR determinant by natural transformation of H. pylori (Wang et al., 1993) on Brucella agar medium containing 18 μg ml−1 of Mtz.

Natural transformation and isolation of spontaneous metronidazole-resistant mutants

Transformation of MtzS strains to MtzR was carried out using a modification of the method of Wang et al. (1993), as follows. Log phase recipient cells (strains 500 or 1134) were prepared in 10 ml of broth from overnight culture in Brucella broth. The bacterial pellet was resuspended in 0.5 ml of TE (Tris EDTA) buffer, and the suspension was spotted onto Brucella agar plates supplemented with 10% FCS. After 3–4 h incubation, 3–8 μg of chromosomal, cosmid or plasmid DNA was spotted onto the bacterial growth followed by incubation for 12–16 h. The bacteria were scraped from the agar surface and suspended in a minimal volume of TE and aliquots were then spread on Brucella agar containing 18 μg ml−1 metronidazole. Transformed colonies were isolated from these plates after 3–4 days' incubation. Spontaneous MtzR mutants were isolated by spreading 0.10 ml of turbid cultures (5 × 109 cells) on Brucella agar containing between 8 and 18 μg ml−1 Mtz.

DNA subcloning and sequencing

Cosmid DNAs that could transform MtzSH. pylori to MtzR were digested with EcoRI; DNA fragments were subcloned into pBluescript plasmid vector by standard techniques (Sambrook et al., 1989), and those able to transform H. pylori to MtzR were identified. In addition, rdxA sequences from various strains of H. pylori were amplified and cloned into pBluescript using primer pairs Mtz6EF (forward) 5′-TGAATTCGAGCATGGGGCAG and reverse primer MtzRBgI 5′-AGCAGGAGCATCAGATAGATCTGA. The resulting amplicons were c. 937 bp in length. DNA sequencing was carried out on both strands manually using the Sequenase kit (Amersham) or by automated methods on a Licor DNA sequencer at the Institute for Marine Biosciences facility of the National Research Council of Canada (Halifax, NS). The sequence was assembled and analysed using the Wisconsin Group GCG software (Devereux et al., 1984) and BLAST search routines to assist identification of ORFs and other sequence features. The sequences have been deposited in GenBank (AF012552, AF012553).

Recombinant rdxA screen for MtzR/S

E. coli DH5α containing pBluescriptSKrdxA clones from all H. pylori strains used in this study were screened for MtzS on Luria Bertani medium containing a range of Mtz concentrations from 0 to 60 μg ml−1. The plates were streaked for isolation of colonies or a 1:100 dilution of a 0.4 OD660 broth culture was spread onto the medium. The plates were incubated under aerobic conditions at 37°C and then scored for growth at 16–24 h.

Allelic exchange mutagenesis and complementation

A 937 bp PCR amplicon of H. pylori MtzS strain H2csr, generated with oligonucleotide primers Mtz6EF and MtzRBgI and cloned into pBluescript-SK, was restricted with SphI, which deleted c. 160 bp fragment from an internal region of rdxA (see Fig. 1 for SphI sites). After gel purification and generation of blunt ends with T4 DNA polymerase, an EcoRV restricted cam cassette originating from Campylobacter coli (Wang and Taylor, 1990) was ligated into rdxA to create pBluescriptrdxA::cam. After transformation into DH5α and plasmid purification, pBluescriptrdxA::cam was introduced into MtzSH. pylori strain 26695 by natural transformation. CmR colonies were picked and then scored for MtzR.

pDH26, a chimeric shuttle vector, was kindly provided by Dr Rainer Haas. H. pylori strain 500 sequences spanning the rdxA ORF were excised from pBluescript by EcoRV and Sal I digestion and subcloned into similarly restricted pDH26. H. pylori strain 1061 was made MtzR by natural transformation of pBluescriptSKrdxA originating from MtzR strain 439. The pDH26rdxA plasmid was introduced into strain 1061MtzR by natural transformation and CmR colonies were scored on BA supplemented with 15 μg ml−1 of Cm. CmR colonies were subsequently screened for MtzS phenotype on Brucella agar containing Cm and 18 μg ml−1 Mtz to demonstrate dominance of wild-type rdxA through loss of the MtzR phenotype.

Enzyme assays

Cell-free extracts were prepared from bacteria that had been grown to mid to late log phase in the appropriate medium and where indicated, either in the presence or absence of 18 μg ml−1 Mtz. The general protocol for preparation of cell-free extracts has been previously described (Hoffman et al., 1996). All enzyme assays were carried out at 25°C in 1 ml volumes in a modified Cary-14 Spectrophotometer equipped with an OLIS data acquisition system (On Line). Nitroreductase activity was assayed with NADH or NADPH at 340 nm (extinction coefficient, 6.22 mM−1 cm−1) or by following the reduction of metronidazole at 320 nm (E = 9.2 mM−1 cm−1). The reaction mixture contained Tris/acetate (100 mM Tris-HCl, 50 mM acetate) pH 7.0, 0.05 mM Mtz and 0.3 mM NADPH or NADH. POR (EC was assayed under anaerobic conditions with 75 mM potassium phosphate (pH 7.3), 10 mM sodium pyruvate, 5 mM benzyl viologen, 0.18 mM coenzyme A (CoA), and 5 μM thiamine PPi as described previously (Hoffman et al., 1996). Reduction of benzyl viologen was followed at 546 nm and specific activity was determined for the reaction using an extinction coefficient of 9.2 mM−1 cm−1. Specific activities were reported as nmoles per min per mg of protein. Protein determinations were performed using the Bradford procedure (Bio-Rad) with bovine serum albumin as the standard.


Many thanks to Paul Sinclair, Jetta Bijlsma and Hans Kusters for helpful discussions, to Louis Bryden for technical assistance and to David Haldane for providing clinical isolates. This work was supported by grants to P.S.H. and S.L.O.Z. from Astra Pharma, Canada and by MRC grant R-14292 and to D.E.B by the US National Institutes for Health AI38166, DK48029, HG00820, TW00611 and the American Cancer Society (VM-121).