INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Helicobacter pylori is a gram-negative microaerophilic bacterium that chronically infects human gastric epithelial cell surfaces and the overlying gastric mucin, a niche that few if any other microbes can occupy. It is carried by more than half of all people worldwide and is an important human pathogen: a major cause of peptic ulcer disease, and a contributor to other illnesses, ranging from childhood malnutrition to gastric cancer, and to increased susceptibility to other food- and water-borne pathogens (7, 8, 32, 38, 47). There is great intrinsic and public health interest in fully elucidating H. pylori's metabolic pathways and how H. pylori maintains its redox balance during microaerobic growth. Such knowledge should help us to understand the extraordinary chronicity of H. pylori infection and factors that determine whether a given infection will be benign or virulent, elucidate mechanisms of drug susceptibility and resistance, and identify potential targets for new effective antimicrobial agents.

Here we focus on mechanisms of susceptibility and resistance of H. pylori to metronidazole (Mtz), a synthetic nitroimidazole that is a key component of popular and affordable anti-H. pylori therapies worldwide and that is also widely used against various anaerobic and parasitic infections (13, 36, 45). Resistance to Mtz is common among H. pylori strains, with frequencies among clinical isolates ranging from 10 to >90%, depending on geographic region and patient group (17, 29, 30). Much of this is attributable to the repeated use of Mtz against other (non-Helicobacter) infections in regimens that are only partially inhibitory, leading to selection for resistance to H. pylori. This is important clinically because Mtz resistance in H. pylori markedly decreases the efficiency of Mtz-based eradication therapy and the cure of associated disease (15, 28).

We had traced the resistance of a Mtz^r clinical isolate to a loss-of-function mutation in rdxA (HP0954), a chromosomal gene for an oxygen-insensitive NADPH nitroreductase, and then identified equivalent rdxA mutations in 15 other Mtz^r strains from North and South America, Australia, and Europe (10, 16). Our experiments also showed that (i) mutational inactivation of rdxA was sufficient to cause Mtz resistance in an Mtz^s reference strain (26695); (ii) expression of rdxA from Mtz^s H. pylori strains in Escherichia coli rendered this normally Mtz^r species Mtz^s; (iii) expression of a functional rdxA⁺ allele on a shuttle plasmid restored Mtz susceptibility to an Mtz^r H. pylori strain; and (iv) new mutations in rdxA, not gene transfer from unrelated lineages, were often responsible for Mtz resistance in clinical isolates (16). In confirmation, rdxA mutations were found in 25 of 27 Mtz^r derivatives of strain SS1 obtained from infected Mtz-treated mice (22) and in 12 of 13 Mtz^r clinical isolates from France and North Africa (42) (the bases of resistance in the unusual Mtz^r strains with apparently intact rdxA genes were not determined). Consideration of enzyme mechanisms had indicated that Mtz activation by the RdxA nitroreductase generates nitroso- and hydroxylamine-related compounds that should be mutagenic and bactericidal (16). Mtz-induced mutation has been documented in H. pylori and also in E. coli carrying an expressed rdxA gene (39). Thus, recurrent exposure of resident H. pylori strains to Mtz, an inadvertent consequence of therapy against other common infections, may induce as well as select for Mtz resistance in this gastric pathogen.

Following our report linking rdxA inactivation and Mtz resistance (16), several researchers suggested that other mechanisms (presumably rdxA independent) might often also cause Mtz resistance (18, 27). This was based in part on observations that nominally Mtz^r clinical isolates differ in the levels of Mtz that they tolerate (resistance level, or MIC), and also fit with precedents of multiple mechanisms of drug resistance in other bacterial species (9, 33, 37). In principle, resistance might also result from (i) diminished Mtz uptake or its active export (26, 40), (ii) more efficient DNA damage repair (6, 43), or (iii) enhanced scavenging of oxygen radicals that are produced according to certain models of Mtz activation (23, 41). Of particular note are plasmid- and transposon-borne nim genes in certain Mtz^r strains of Bacteroides fragilis that promote conversion of nitroimidazoles from prodrug to harmless amino derivatives, rather than to toxic nitroso radicals, and that thus confer resistance without loss of chromosomal nitroreductase gene function (5, 46).

DNA fingerprint and sequence analyses have indicated that each H. pylori clinical isolate differs genetically from most other independent isolates (1, 2). Superimposed on this great general diversity, we and others have identified several subpopulations of H. pylori that are relatively distinct genetically, with each specific to a different geographic region or human ethnic group: one in southwest Europe (Spain); a second in East Asia; and a third in Calcutta, India (1, 21, 24, 30, 31). The strains of South and Central America seemed most closely related to southwest European (Spanish) strains, not Asian strains, as are many strains from Africa and the United States (24). Most or all strains whose Mtz resistance has been studied to date are probably of the European type; the possibility of alternative resistance genes being abundant in the gene pools of non-Western H. pylori strains remains to be tested.

Here we describe functional and sequence analyses that (i) establish that mutational inactivation of the rdxA nitroreductase gene is critically involved in primary Mtz resistance in most or all strains from South and East Asia and sub-Saharan Africa, as well as from the West; (ii) demonstrate that the resistance of Mtz^r (rdxA-deficient) strains can be increased by mutation in other genes, including frxA (flavin nitroreductase; HP0642 in reference 42); and (iii) show that frxA does not contribute to the normal Mtz^s phenotype of wild-type H. pylori strains. No evidence of determinants that bypass the need for rdxA inactivation in the development of clinically significant Mtz resistance was found.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and culture conditions. The H. pylori strains used in this study were clinical isolates from diverse parts of the world, and most have been described in references 24 and 30. The recent clinical isolates from Alaska are from Native peoples in Anchorage and other sites around the state. Strains 26695 and J99 are Mtz^s reference strains whose complete genome DNA sequences have been determined (3, 44). Each strain is from an ethnic European patient: 26695 from the United Kingdom (44), and J99 from an ethnic European in Pulaski, Tenn. (T. L. Cover, personal communication).

DNA methods. H. pylori genomic DNAs were isolated from confluent cultures grown on BHI agar using a Qiamp tissue kit (Qiagen Corporation, Chatsworth, Calif.) or the cetyltrimethylammonium bromide-phenol method (4). PCR was carried out in 20-µl volumes containing 10 ng of genomic DNA, 10 pmol of each primer, 1 U of Taq DNA polymerase (Promega) or high-fidelity Taq (Boehringer Mannheim), and 0.25 mmol of each deoxynucleoside triphosphate in standard PCR buffer. Reaction mixtures were preincubated for 2 min at 94°C and then used in 30 cycles of 94°C for 40 s, 58°C for 40 s, and 72°C for 1 min per kb, with the final elongation step of 72°C for 10 min. PCR fragments were purified for sequencing by QIAquik PCR purification kit (Qiagen). Sequencing reactions were carried out using a Big Dye terminator cycle sequencing kit (PE Applied BioSystems, Foster City, Calif.), and products were run on ABI automated sequencers in the Washington University molecular microbiology core facility. The primers used are listed in Table 1.

Determination of Mtz sensitivity and resistance. Frozen H. pylori cultures were streaked onto Mtz-free BHI agar and incubated for 3 days; then bacterial growth was respread on fresh Mtz-free BHI agar and incubated for 1 day. The resulting young exponentially growing cells were suspended in phosphate-buffered saline; a series of 10-fold dilutions of these suspensions was then prepared, and 10 µl of each dilution was spotted on freshly prepared BHI agar containing appropriate concentrations of Mtz (variously, 0, 0.2, 0.5, 1.5, 3, 8, 16, 32, and 64 µg/ml). When the frequency of cells that formed colonies on Mtz-containing media was very low (<10⁶), estimates of viability and mutant frequency were made more accurate by spreading aliquots of cultures on the entire surface of a BHI agar petri plate, instead of spotting aliquots in small areas. A strain was considered to be susceptible to concentrations of Mtz that decreased its efficiency of colony formation at least 10-fold. This quantitative procedure was more sensitive than conventional MIC determinations, which typically estimate concentrations of antibiotic needed to block growth of denser bacterial suspensions. In particular, this procedure minimizes complications that could stem from the mutagenicity of Mtz for H. pylori (39), which would be exacerbated if H. pylori stressed by DNA-damaging agents tended to enter a hypermutable state (35).

New Mtz^r mutants. Mtz^r mutant derivatives of strain 26695 that may have induced as well as selected by Mtz (39) were obtained by spreading 10⁸ bacterial cells from young cultures (as above) on BHI agar containing Mtz at 3 µg/ml. Individual colonies were streaked on BHI agar with the same Mtz concentration and then tested on BHI agar containing higher concentrations of Mtz (8, 16, 32, and 64 µg/ml).

Engineered H. pylori strains. (i) rdxA mutants. The rdxA::cam allele, which contains a cam cassette in the rdxA gene (16), was moved to the chromosome of H. pylori strains by DNA transformation and selection on BHI agar with 15 µg of chloramphenicol (Cam)/ml. Transformants were checked to verify that they had resulted from allelic replacement by PCR using primers rdxA-F and rdxA-R, which generates products of 2 kb in cases of replacement and 886 bp in cases of retention of the original rdxA allele (Table 1).

(ii) frxA gene cloning and construction of frxA::cam and frxA::kan insertion/deletion mutants. The frxA flavin nitroreductase gene segment (HP0642 in the strain 26695 genome [44]) was PCR amplified from H. pylori genomic DNAs using primers frxA-F1 and frxA-R1 or, in some experiments, frxA-F and frxA-R. The amplified DNAs were cloned into the EcoRV site of plasmid pBS and recovered after transformation of E. coli DH5.

(iii) HP1508::cam insertion/deletion mutation. The HP1508 gene, which encodes a ferredoxin-like protein of unknown function, was PCR amplified from strain 26695 DNA, using primers 1508-F1 and 1508-R1, and similarly cloned into pBS. A marked 1,108-bp internal deletion was generated using PCR with outward-facing primers 1508-F2 and 1508-R2, followed by ligation with a cam cassette, as above. The structures of Cam^r transformants were verified by PCR with primers 1508-F1 and 1508-R1.

(iv) oorD::cam insertion/deletion mutation. The oorD (HP0588) gene, which encodes the ferridoxin component of the multisubunit oxoglutarate oxidoreductase enzyme (20), was PCR amplified from strain 26695 DNA, using primers 588-F1 and 588-R1, and cloned into pBS. A marked 537-bp internal deletion was generated using PCR with outward facing primers 588-F2 and 588-R2, followed by ligation with a cam cassette, as above. The structure of the one H. pylori transformant obtained (see Results) was verified by PCR with primers 588-F1 and 588-R1.

(v) Addition of a functional rdxA gene to the H. pylori genome. A functional rdxA gene, PCR amplified from strain 26695 with primers rdxA-F1 and rdxA-R1, and the cam cassette were cloned into the EcoRV and SmaI sites, respectively, of plasmid pBS (each gene transcribed toward the other). In parallel, a 1.37-kb segment of cagA, PCR amplified from strain NCTC11637 using primers cagA2143F and cagA3512R, was cloned into the SmaI site of pBS. The resulting plasmid was linearized by PCR with outward-facing primers cagA2673R and cagA2900F, which are specific to sites 531 and 613 bp from the two ends of the cloned cagA DNA fragment. The rdxA-cam segment was then PCR amplified from the pBS-rdxA-cam construct, using pBS-specific primers M13F and M13R, and cloned between the 531- and 613-bp fragments of the cagA gene. An isolate in which rdxA was oriented in the same direction as cagA was used to transform NCTC11637 to Cam^r. The structures of resultant transformants were verified by PCR using primers cagA2143F and cagA3512R.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Intrinsic Mtz susceptibility or resistance of H. pylori reference strains and clinical isolates. To determine the lowest concentrations of Mtz that permitted survival of reference strains 26695 and J99, young cultures were diluted serially, aliquots of dilutions were spotted on Mtz-containing BHI agar, and numbers of colonies formed at appropriate dilutions were determined. Strain 26695 exhibited an efficiency of plating (EOP) of ~1 on BHI agar with up to 1.5 µg of Mtz/ml and ~10⁴ on BHI agar with 3 or 8 µg of Mtz/ml (phenotype designated 1.5R 3S). Reference strain J99 was somewhat more susceptible, exhibiting EOPs of ~1 and 10³ on BHI agar with 1 and 1.5 µg of Mtz/ml, respectively (phenotype designated 1R 1.5S) and 10⁴ on BHI agar with 3 or 8 µg of Mtz/ml (Fig. 1).

View larger version (45K): [in this window] [in a new window]   FIG. 1.   Profiles of intrinsic susceptibility and resistance to Mtz of reference strains 26695 and J99. Young exponentially growing cultures were diluted, aliquots were spotted on BHI agar with indicated concentrations of Mtz, and surviving colonies were counted from appropriate dilutions, as detailed in Materials and Methods. Presented are the average and range of results with two single colony isolates (cultures) of each strain, with assays repeated three times with each culture.

New Mtz^r mutants generated in culture. To test our inference that Mtz resistance generally involves decreased rdxA function, new mutant Mtz^r derivatives of reference strain 26695 were selected on BHI agar containing just 3 µg/ml, the lowest concentration of Mtz that allowed Mtz^r mutants to emerge cleanly from background growth. Such mutants were obtained at frequencies of about 10⁴ in cultures from different single-colony isolates, as noted above (Fig. 1).

Nature of newly arisen Mtz^r mutants. Eight low-level Mtz^r mutants (3R 8S) were characterized by sequencing. Three contained mutations in or immediately upstream of rdxA, whereas the other five did not (Table 4, group A); the mutations that caused the weak Mtz resistance of these latter five isolates have not been identified. The rdxA genes of four independent Mtz^r mutants with the more common higher-level Mtz resistance were also sequenced (two 8R 16S, one 16R 32S, and one 32R 64S). Simple point mutations in rdxA were found in each case (Table 4, group B), as expected (16).

Loss-of-function mutations in rdxA associated with Mtz^r worldwide. The idea that rdxA inactivation is critically involved in most or all clinically significant cases of Mtz resistance was tested against an alternative possibility, that Mtz resistance in certain geographic regions might often result from auxiliary (e.g., plasmid or transposon) resistance genes that are uncommon in Western H. pylori strains and that bypass the need for rdxA inactivation. This entailed PCR amplification of a segment containing rdxA from Mtz^r strains from various representative populations (12 Chinese, 12 Indian, 11 Alaska Native, 9 Peru Native, and 6 South African), electroporation of the Mtz^s strain 26695 with these PCR-amplified rdxA DNAs, and quantitation of the yield of Mtz^r transformants on BHI agar with 8 µg of Mtz/ml (Fig. 2). Mtz^r transformants were obtained with frequencies of about 10², using rdxA DNA amplified from each of the 50 Mtz^r clinical isolates tested and also from strain SS1rdxA111, as a positive control. This frequency was 100-fold higher than the yield of Mtz^r colonies obtained with PCR products from each of six control Mtz^s strains (~10⁴) (two Alaska Native, two South African, and two Indian), indicating that each of the 50 Mtz^r strains contained mutant alleles of rdxA. It is also noteworthy that each of the 50 rdxA mutant PCR products was of the size expected (~890 bp), indicating that each contained a point mutation, not an insertion or deletion, in rdxA. These results showed that rdxA inactivation is critically involved in most cases of Mtz resistance worldwide and ruled out a model (14) in which changes in regulatory genes that affect expression of rdxA and/or other reductase genes would be responsible for most Mtz resistance in clinical isolates.

View larger version (15K): [in this window] [in a new window]   FIG. 2.   DNA transformation strategy for testing involvement of rdxA gene mutation in Mtzr clinical isolates.

RdxA as principal determinant of Mtz susceptibility of most wild-type strains. Initial tests had shown that transformants of the Mtz^s strain 26695 obtained using an rdxA::cam cassette and selected for Cam^r were Mtz^r in phenotype (16). In the present, more quantitative tests, this rdxA::cam mutant strain exhibited a 16R 32S phenotype, as did most newly arisen Mtz^r mutants. Equivalent 16R 32S phenotypes were exhibited by derivatives of strain 26695 containing unmarked rdxA deletion alleles (rdxA601 or rdxA111) (Fig. 3), in each case selected after DNA transformation and selection for resistance to just 3 µg of Mtz/ml. These results showed that rdxA encodes the only nitroreductase sufficiently active to confer an Mtz^s phenotype on this reference strain.

View larger version (47K): [in this window] [in a new window]   FIG. 3.   Mtz resistance profiles of rdxA111 or rdxA601 deletion mutant transformants of strain 26695 and of rdxA601-containing transformants of three 26695 derivatives that had exhibited an unusual 3R 8S phenotype that was not attributable to mutation in rdxA (201, 116, and 619 [Table 4]). Transformants containing the rdxA111 and rdxA601 alleles were selected using BHI agar with 8 µg of Mtz/ml. Presented are the average and range of results with three single colony isolates (cultures) of each strain, with each culture assayed two times.

Mutations in other genes can affect the level of Mtz resistance. Three sets of results established that the level of resistance of a typical Mtz^r (rdxA-deficient) H. pylori strain can be affected by other genetic determinants. First, introduction of the rdxA601 deletion into three derivatives of strain 26695 with weak Mtz^r phenotypes (3R 8S) that were ascribed to unknown sequence changes outside rdxA (see above) resulted in 16R 32S phenotypes in each case. However, the EOP of these transformants on BHI agar with 32 µg of Mtz/ml was reproducibly ~10²that is, several hundred-fold higher than that of control rdxA601 transformants of 26695 wild type, selected in parallel (Fig. 3). This suggests that the mutations responsible for the 3R 8S phenotype may have affected process(es) distinct from those controlled by RdxA nitroreductase itself.

4

7

9

View larger version (44K): [in this window] [in a new window]   FIG. 4.   Mtz resistance profiles of the unusual highly resistant mutant of strain 26695 (designated 26695 mutant 161 [Table 4]) and also a 26695 derivative that contains only the rdxA gene from this mutant (designated rdxA.161). The rdxA.161 transformant was selected on BHI agar with 8 µg of Mtz/ml. The presence of the expected GAG-to-TAG change in rdxA (Table 4) was verified by DNA sequencing. Presented are the average and range of results with five single colony isolates (cultures) from each strain, with each culture assayed once.

Flavin nitroreductase (frxA gene product) contributes to residual Mtz susceptibility of rdxA mutant strains. Theoretical considerations had suggested that frxA [HP0642, encoding NAD(P)H-flavin oxidoreductase; an rdxA paralog] might contribute to Mtz susceptibility in H. pylori (16). Although one frxA gene clone did not make E. coli susceptible to Mtz (16), further studies identified an frxA-containing cosmid clone from an H. pylori strain with a high-level Mtz^r phenotype (32R) (strain 439 in reference 16) that increased the yield of Mtz^r H. pylori transformant colonies when mixed with rdxA mutant DNA but did not transform Mtz^s H. pylori to Mtz^r when used alone. Given these various results, we elected to reexamine the possibility of a role for frxA in Mtz susceptibility and resistance. First, we sought to again PCR amplify and clone frxA-containing DNA segments from several different Mtz^s H. pylori strains, but using a high-fidelity Taq polymerase formulation to minimize mutation during PCR. Four of ten independent frxA-containing pBS plasmid clones that were recovered in E. coli DH5 (two from 26695; one each from SS1 and HP500) resulted in susceptibility to Mtz. In quantitative determinations using frxA clones from 26695, the EOPs were about 0.01 and 0.001 on L agar with 1 and 2.5 µg of Mtz/ml, respectively, whereas the parental E. coli strain (lacking frxA) exhibited an EOP of 1 on L agar with 50 µg of Mtz/ml. It was also noted that E. coli carrying cloned functional frxA genes tended to make small colonies on Mtz-free L agar. This result suggested that the poor yield of frxA-containing clones that rendered E. coli Mtz^s (only 4 of 10) might be due to some toxicity of frxA when hyperexpressed and that the initial lack of Mtz susceptibility associated with frxA cloning (16) was a spurious result, perhaps reflecting mutation during PCR or cloning and unwitting selection of a healthy (frxA mutant) transformant colony.

View larger version (42K): [in this window] [in a new window]   FIG. 5.   Effect of frxA inactivation on Mtz susceptibility and resistance depends on whether rdxA is functional or not. (A) Profile of Mtz susceptibility of an frxA-null mutant of an rdxA::cam strain and of its frxA+ parent; (B) profile of Mtz susceptibility of an frxA-null mutant derivative of 26695 (rdxA+) and of its wild-type parent. Presented are the average and range of results with three single colony isolate cultures of each strain, with each culture assayed two times.

Other possible contributors to resistance. As noted above, the 26695 rdxA frxA double mutant had a 32R 64S phenotype. Derivatives with higher resistance (64 instead of 32 µg of Mtz/ml) were recovered at a frequency of ~10⁴, in contrast to ~10⁷ in the case of frxA⁺ rdxA-deficient strains (Fig. 5A). This indicated that resistance can be enhanced by mutation in at least one additional locus.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We have studied mechanisms of susceptibility and resistance to Mtz in H. pylori strains from diverse parts of the world, motivated in part by recent findings that different H. pylori genotypes predominate in East Asia, South Asia, and Europe, and that Latin American and African and many U.S. strains tend to be most closely related to those of Europe, not Asia (24, 30, 31). Here we present mutational, gene cloning, and sequence analyses that confirm and extend our initial conclusions (10, 16) in establishing (i) that the lethality of Mtz to wild-type H. pylori depends primarily on the activity of an oxygen-insensitive NADPH nitroreductase encoded by the rdxA (HP0954) gene, which mediates conversion of Mtz from harmless prodrug to toxic and mutagenic product (16, 39); and (ii) that Mtz resistance generally results, at least in part, from mutations that inactivate rdxA.

In the present studies of East Asian (China and Japan), South Asian (Calcutta), South African, and Alaska Native strains, as well as Western (Spain and Amerindian Peruvian) strains, we found (i) that a functional rdxA nitroreductase gene is primarily responsible for the high susceptibility to Mtz of most or all wild-type H. pylori strains; (ii) that clinically significant Mtz resistance generally requires mutation in rdxA; and (iii) that the level of Mtz resistance that a strain exhibits can be further enhanced by additional changes elsewhere in its genome, but only if it is already mutant in rdxA. With only a few possible exceptions (discussed below), no evidence of auxiliary resistance genes that confer clinically significant Mtz resistance without rdxA inactivation was found in any population. This is noteworthy, because many of the strains examined came from societies in which H. pylori infection and Mtz usage are frequentconditions that would have favored the spread of any plasmid- or transposon-borne auxiliary resistance determinants.

That additional genes might also be important is emphasized by the finding that more than half of Mtz^r clinical isolates were resistant to levels of Mtz higher than can be obtained by inactivation of rdxA alone (Table 3). Further analyses indicated that this enhanced resistance can occur stepwise, by mutation in at least two other loci in strains already mutant in rdxA. First, mutational inactivation of frxA (HP0642), an rdxA homolog that encodes a related reductase (24% amino acid sequence identity), increased the resistance of rdxA-deficient H. pylori from 16 to 32 µg/ml. However, frxA inactivation, by itself, had little effect on the intrinsic Mtz susceptibility of Mtz^s strains. This is in accord with evidence that rdxA inactivation is sufficient to render Mtz^s strains Mtz^r. In this context, our finding that cloned frxA⁺ genes from each of several Mtz^s H. pylori strains made E. coli highly susceptible to Mtz suggests that synthesis or activity of the FrxA reductase may be down-regulated in H. pylori. In accord with this view, we found two unusual clinical isolates that became Mtz^r only if rdxA and frxA were each inactivated. In parallel studies, we have also found that the special mouse-adapted strain SS1 also requires inactivation of both frxA and rdxA to achieve an Mtz^r phenotype and that frxA mRNA levels in this strain are higher than in reference strains (J. Y. Jeong and D. E. Berg, unpublished data). We have begun to search for the putative regulatory gene(s) and/or site(s) that may be mutant in these strains and to test the possibility that unusually high FrxA activity may contribute to bacterial fitness in certain hosts.

Even higher-level resistance (64R phenotype) is common among clinical isolates and is readily obtained in culture, starting with an rdxA frxA double-mutant strain. Hence, at least one other gene must be involved in residual Mtz susceptibility and the emergence of hyperresistance. The involvement of other reductase enzymes in susceptibility is suggested by our finding that Mtz can be mutagenic even for hyperresistant H. pylori strains, since Mtz-promoted mutagenesis reflects enzymatic reduction of Mtz (39). The gene responsible for this third incremental component of hyperresistance has not yet been defined.

The multiplicity of metabolic and housekeeping functions that potentially can affect Mtz susceptibility and resistance is further illustrated by our finding that five of the eight derivatives of strain 26695, selected for very slight decreases in susceptibility to Mtz (3R instead of 1.5R phenotype), had resulted from mutation outside of rdxA. One explanation invokes polar mutations in upstream sequences that simply decrease rdxA expression. This explanation seems unlikely, however, since the level of Mtz resistance achieved after transformation of these mutants with an rdxA deletion allele was slightly but reproducibly higher than in their isogenic parent (Fig. 3). This implies that the 3R 8S and rdxA mutations affect quite different processes or pathways. The mutations are also unlikely to be in frxA: their enhancement of Mtz resistance in strain 26695 wild type, although slight, was greater than that conferred by an frxA-null allele, whereas they had less effect on Mtz resistance than the frxA-null allele in the rdxA-null background. Given the mutagenic and DNA-damaging effects of products of Mtz activation (39) and the dramatic increase in Mtz susceptibility caused by recA gene inactivation (43), these subtle mutant phenotypes might be ascribed to changes in genes affecting DNA replication or repair (6, 43), or equally to changes in efficiency of Mtz uptake (26) or efficiency of physiologic adaptation to growth with Mtz (19).

Also meriting further study are a few exceptional Mtz^r strains that were reported by others to contain normal rdxA sequences: 2 of 27 Mtz^r variants recovered from mice infected with strain SS1 and treated subtherapeutically with Mtz (22), and 1 of 13 Mtz^r strains from France and North Africa (42). It should now be possible to learn if any of these unusual mutants have decreased RdxA synthesis or activity, or if any of them result from an alternative, but still rare, mechanism for Mtz resistance that bypasses the need for rdxA inactivation.

The distribution of various levels of Mtz resistance among clinical isolates differs markedly from that obtained by one-step forward mutation to Mtz^r in culture. We propose that this distribution reflects a complex dynamic, including (i) the mutagenic effects of Mtz activation; (ii) the intensity of selection for Mtz^r phenotypes during Mtz-based therapy, which is dictated by amounts of Mtz administered, frequency and duration of treatment, and gastric acidity or physiologic parameters that affect drug potency in H. pylori's mucosal niche; and (iii) possible effects of resistance on H. pylori fitness during periods between therapy. Given the diversity among H. pylori strains and their human hosts, the evolutionary cost of a given level of Mtz resistance may depend on various aspects of bacterial genotype that affect the overall flow of metabolites during growth, and also on aspects of human genotype and physiology that affect human susceptibility to a given H. pylori strain (11, 12). Many of these issues should soon be clarified through high-resolution H. pylori molecular genetics and use of appropriate in vitro culture strategies and well-chosen experimental animal infection models.