INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The cultivation of Helicobacter pylori from the human gastric mucosa in the early 1980s (33) and the demonstration of its relationship to gastritis, peptic ulcer disease, and gastric neoplasia triggered a series of studies of the ecology and role of Helicobacter-like organisms in a variety of hosts. At present, 18 validly named species and two candidate species (3, 4) have been isolated from gastric or enteric samples of a variety of hosts, including humans, hamsters, rats, mice, ferrets, pigs, poultry, wild birds, cats, dogs, monkeys, and cheetahs. Several other Helicobacter taxa remain unnamed or have not been properly described in accordance with internationally accepted rules of nomenclature, and therefore their names have not been validated (6, 9, 13, 29, 30). The biochemical inertness of all Campylobacter-like organisms, including helicobacters, plays a major role in influencing the identification strategies of clinical laboratories. Classical phenotypic tests routinely used for the identification of clinical bacteria often yield negative or variable results within species. Problems associated with phenotypic identification have led to sequence analysis of rRNA genes (in particular 16S) as an increasingly popular alternative approach for identification of new isolates. Strains have been identified as novel species, primarily because of supposedly sufficient differences in 16S ribosomal DNA (rDNA) sequence similarity to known species (for example, reference 13), or as well-established species, again primarily based on the percentage of similarity of 16S rDNA sequence (for example, references 7 and 28).

In the present study, we describe the identification and characterization of 14 Helicobacter cinaedi isolates obtained from various hosts, including humans, dogs, foxes, and a rat, by using different phenotypic and genotypic approaches and comment on the pitfalls of the 16S rDNA sequence approach. These 14 strains included isolates described in this journal as "Helicobacter westmeadii" sp. nov. (30) and as Helicobacter sp. nov. strain Mainz (13).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. Three H. cinaedi reference strains (LMG 7543^T, LMG 8770, and LMG 9071) from the original study by Totten et al. (29) and five additional strains (LMG 8558, LMG 8559, LMG 9072, LMG 9153, and LMG 9357), characterized in a previous polyphasic taxonomic study (31), were used as references (Table 1). Fourteen recent isolates obtained from the feces of dogs, foxes, and a rat and from human blood or feces were studied; the isolates were from Finland (n = 6), Belgium (n = 2), Sweden (n = 2), Australia (n = 2), Scotland (n = 1), and Germany (n = 1) (Table 1).

PAGE of whole-cell proteins. All 14 isolates were grown on Mueller-Hinton agar (catalog no. CM 337; Oxoid, Ltd., Basingstoke, United Kingdom) supplemented with 5% (vol/vol) horse blood and were incubated at 36 to 37°C in a microaerobic atmosphere containing approximately 5% O₂, 3.5% CO₂, 7.5% H₂, and 84% N₂. Whole-cell protein extracts were prepared, and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (PAGE) was performed as described before (26). Whole-cell protein profiles of H. cinaedi reference strains and of type and reference strains of other Helicobacter species were available from previous studies. The densitometric analysis, normalization, and interpolation of the protein profiles and numerical analysis were performed with the GelCompar software package version 4.2 (Applied Maths, Kortrijk, Belgium). The profiles were recorded and stored on an IBM PC-compatible computer. The similarities between all pairs of traces were expressed by the Pearson product moment correlation coefficient presented below as percentages of similarity for convenience.

Phenotypic analysis. All 14 isolates and reference strain LMG 9357 (not examined before by this scheme) were grown on 5% (vol/vol) calf blood agar for 3 days under microaerobic conditions, as described previously (19). A total of 64 phenotypic characters were determined using methods described previously (19-22). The results were compared with data for 37 Campylobacter taxa in a probability matrix using computer-assisted methods.

Dot blot DNA-DNA hybridizations. DNA was isolated by the method of Pitcher et al. (25), with modifications described before (14). Dot blot DNA-DNA hybridizations were performed as described before (10, 14). The DNA of H. cinaedi LMG 7543^T was used as a probe. DNAs of the following strains were hybridized with the LMG 7543^T probe: LMG 7543^T, LMG 8559, and LMG 8558 (the three reference strains) and R-927, R-2971, R-2977, R-2981, R-2983, R-2991, and R-5759. "Flexispira rappini" CCUG 23435 and Helicobacter canis CCUG 19561 served as negative controls. Aliquots of 0.5, 5, and 50 ng of chromosomal DNA were spotted onto the membranes for each hybridization experiment.

16S rDNA sequencing. The primers and methods used for DNA extraction, PCR amplification, and direct, automated sequencing of 16S rRNA genes were as described previously (18), except that primer 1492r (5'-TACGGYTACCTTGTTACGACTT) was used in place of 1392r for the initial PCR amplification and subsequent sequencing of the PCR product. This allowed sequences to be obtained (both strands) over ~95% (compared to ~90%) of the 16S rRNA gene. In addition, the Taq polymerase and PCR buffer (1.5 mM final MgCl₂ concentration) were from Boehringer (Mannheim, Germany). The consensus sequence and the sequences of strains belonging to the same phylogenetic group (retrieved from the EMBL data library) were aligned, and a phylogenetic tree was constructed based on the neighbor-joining method by using the GeneCompar version 2.0 software package (Applied Maths).

Nucleotide sequence accession numbers. The nucleotide sequence accession numbers for the 16S rDNA sequences of strains R-4792, LMG 16312, and R-927 are AF207737, AF207738, and AF207739, respectively.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Whole-cell protein electrophoresis. All 14 isolates were identified as H. cinaedi by comparison of their whole-cell protein profiles with a database comprising patterns of over 1,000 Campylobacter, Arcobacter, Helicobacter, and Wolinella strains. Figure 1 shows the result of the numerical comparison of the protein patterns of the 14 isolates, the eight H. cinaedi reference strains, and the type or reference strains of other Helicobacter species. All 22 H. cinaedi strains form a single protein electrophoretic cluster above a similarity level of 80%.

View larger version (39K): [in this window] [in a new window]   FIG. 1.   Dendrogram expressing similarity of whole-cell protein patterns of all 22 H. cinaedi isolates and of Helicobacter reference strains. The strain numbers in boldface indicate those strains that were included in DNA-DNA hybridization experiments. The strain marked with an asterisk was originally described as Helicobacter sp. strain Mainz; the strains marked with double asterisks were originally described as "H. westmeadii".

Phenotypic analysis. Computer-assisted comparison of the biochemical profiles identified 14 of the 15 strains examined (the 14 novel isolates and reference strain LMG 9537) as H. cinaedi with a Willcox probability identification score (IDS) of 0.99 (i.e., 99% probable) or more. Field strain LMG 16312 was not confidently identified by this method (IDS, <0.95). Several atypical features (notably results in nitrate reduction, alkaline phosphatase production, indoxyl acetate hydrolysis, tolerance and reduction of triphenyl tetrazolium chloride, growth at 42°C, and resistance to cephalothin and cefoperazone) were noted in the phenotype of LMG 16312 compared with those of type and reference strains determined previously (23). Table 2 summarizes the key phenotypic features of the H. cinaedi strains examined here and previously (23) that are especially useful for discriminating among other enteric helicobacters associated with human gastroenteritis or septicemia.

Dot blot DNA-DNA hybridizations. Dot blot DNA-DNA hybridization experiments revealed strong hybridization reactions between the H. cinaedi LMG 7543^T probe and all of the H. cinaedi strains tested. Weak or no hybridization was obtained to "F. rappini" or H. canis DNA (the nearest phylogenetic neighbors of H. cinaedi) or to any other Helicobacter species (Fig. 2).

View larger version (58K): [in this window] [in a new window]   FIG. 2.   Dot blot DNA-DNA hybridization result obtained with the labeled H. cinaedi LMG 7543T probe. The strain marked with an asterisk was originally described as Helicobacter sp. strain Mainz; the strain marked with double asterisks was originally described as "H. westmeadii".

16S rDNA sequencing and phylogenetic analysis. The expected (~1,500-bp) amplicon was obtained from both R-4792 and R-927, whereas strain LMG 16312 yielded an enlarged amplicon of ~1,700 bp, later found to be due to the presence of an intervening sequence (IVS) of ~200 bp that started at base 168 of the submitted sequence (AF207738) for this strain. The 16S rDNA sequence similarity levels between the different H. cinaedi isolates varied considerably. Figure 3 summarizes the 62 polymorphic sites (4.3%) identified after alignment of our H. cinaedi sequences with an existing sequence for strain R-927 (accession number X81028) and the 1,444-bp sequence of the H. cinaedi type strain (accession number M88150). The majority of these polymorphic sites were accounted for by the two sequences representing strain R-927. The positions where alignment gaps were introduced into each sequence are also shown in Fig. 3. Three of these seven alignment gaps occurred at the two termini of the IVS for strain LMG 16312 (the IVS sequence itself is not shown), and a single-base deletion was also found in this region for strain R-927 (both sequences). The remaining alignment gaps seem likely to be errors in downloaded EMBL sequences, since the 1-base deletion (position 184) and 2-base insertion (position 826) were not found in any of our submitted sequences. The repeat sequence for R-927 was highly similar (99.9%) to the sequence originally described (13), except that a 2-base insertion (detailed above) was seen in the original sequence and a further four polymorphisms were identified (positions 928 and 1419 to 1421 [Fig. 3]), possibly as a result of nucleotide misincorporation during sequencing procedures in either of the two investigations.

View larger version (21K): [in this window] [in a new window]   FIG. 3.   Comparison of H. cinaedi 16S rRNA gene sequences with the sequence for the type strain (M88150), showing only polymorphic sites and those sites where gaps () were introduced during alignment. X81028 represents a previously published sequence for strain R-927 (13). The numbers (in vertical format) above the sequences correspond to the nucleotide numbering of the M88150 sequence, including three alignment gaps. Bases identical to the M88150 sequence are represented by periods.

View larger version (26K): [in this window] [in a new window]   FIG. 4.   Neighbor-joining phylogenetic tree of H. cinaedi strains and related bacteria based on 16S rRNA sequence comparisons. The scale bar indicates 5% sequence dissimilarity.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

It is well known and widely accepted that the biochemical identification of Helicobacter strains by means of a limited number of commonly used tests is extremely difficult. In the present study, we used two conceptually different approaches to identify 14 putative helicobacters to the species level. The accuracy of the identifications suggested by these methods was checked by DNA-DNA hybridization, which is generally considered the standard for species discrimination (34). Previous studies with a range of different bacteria have demonstrated a correlation between the degree of whole-cell protein pattern similarity as revealed by one-dimensional sodium dodecyl sulfate-PAGE and the level of DNA-DNA hybridization. Vandamme et al. (31) and Costas et al. (2) previously demonstrated that protein pattern analysis is most useful to distinguish Helicobacter species. This was recently confirmed by Jalava et al. (15), who compared DNA-DNA hybridization, whole-cell protein electrophoresis, and biochemical analysis for the differentiation of several gastric Helicobacter species. Similarly, the application of biochemical examination by means of a panel of 64 tests for the identification of Helicobacter strains was validated by the examination of a large collection of well-characterized strains (1, 23).

In the present study, whole-cell protein pattern analysis identified all 14 field isolates unambiguously as H. cinaedi, while the probabilistic analysis of over 60 biochemical test results successfully identified 13 of these strains, as well as a reference strain not examined before. A qualitative DNA-DNA hybridization assay of a selection of 10 strains confirmed that all of the strains belonged to a single species. Previous comparative studies of this qualitative DNA-DNA hybridization assay and the quantitative optical renaturation rates method (5) revealed that the former procedure allowed correct species level classification of strains of various Helicobacter species (14).

In contrast, identification of several of these isolates based on a restricted set of biochemical tests and their levels of 16S rRNA sequence similarity to known Helicobacter species was not straightforward. Husmann et al. (13) reported that strain R-927 represented a novel Helicobacter species referred to as Helicobacter sp. strain Mainz. In our study, its biochemical reactivity pattern and whole-cell protein profile conformed to those of the other H. cinaedi strains (Fig. 1), and DNA-DNA hybridization to the type strain of H. cinaedi revealed a very strong signal (Fig. 2). The main reason for the misidentification of this strain was an overreliance on its unique position in the 16S rDNA sequence-based phylogenetic tree (13) (Fig. 4). This strain indeed has a 16S rRNA gene sequence that is strikingly different from those of other H. cinaedi strains and is in fact more similar to that of the H. fennelliae type strain (Fig. 4). It must be emphasized here that divergence in 16S rRNA gene sequences of up to 4.3% have been reported in another bacterium belonging to the epsilon subdivision of the division Proteobacteria, Campylobacter hyointestinalis (11). For that species, too, DNA-DNA hybridizations confirmed that these strains represented a single genomic species as defined in taxonomic practice (34). For species like this, it is not surprising to find that not all strains cluster together in a phylogenetic tree (Fig. 4).

The classification of strains R-5758 and R-5759 as a novel Helicobacter species, "H. westmeadii" (30), was probably primarily due to failure to grow the isolates under optimal conditions, leading to at least a few equivocal biochemical test results. "H. westmeadii" was reported to be an anaerobic helicobacter, thereby differentiating it from H. cinaedi, its nearest phylogenetic neighbor (Fig. 4). However, like many other helicobacters, both isolates indeed grew very poorly in a microaerobic environment without hydrogen but grew abundantly when hydrogen was supplemented. Incubation of the strains in an incubator routinely used to work with strict anaerobes (atmospheric composition, 5% CO₂, 10% H₂, and 85% N₂) did not yield visible growth. Further biochemical tests performed under these optimal conditions failed to reproduce hippurate activity, and the overall reactivity pattern corresponded to that of typical H. cinaedi strains. Again, DNA-DNA hybridization to the type strain of H. cinaedi revealed a very strong signal (Fig. 2), confirming the identification of these isolates as H. cinaedi. Therefore, the combined evidence from biochemical, protein electrophoretic, and DNA-DNA hybridization analyses indicate unambiguously that "H. westmeadii" is a junior synonym of H. cinaedi.

Undoubtedly, comparison of (nearly) entire 16S rDNA sequences is one of the most powerful tools for establishing the phylogenetic neighborhood of an unknown organism. However, many taxonomic studies have revealed that this approach is often not sensitive enough to identify strains to the species level. Indeed, strains belonging to different species may have identical 16S rRNA gene sequences, and strains of one species may have 16S rRNA genes that differ by up to 3% (27) and even over 4% (11) of the total 16S rRNA gene sequence. There is clearly a lack of knowledge, not only of the strain-to-strain variation within a species, but also of the interoperon variation within a single strain. Therefore, concluding that an unidentified isolate belongs to a particular Helicobacter species because it shares a high percentage of its 16S rRNA gene sequence or concluding that it represents a novel species because it occupies a unique position in the phylogenetic tree or because it shares only 97% of its 16S rRNA gene sequence with its closest neighbor is premature in the absence of appropriate complementary data.

The identification of the present collection of strains as H. cinaedi expands the number of potential reservoirs for infection. H. cinaedi was first described by Totten et al. (29) to delineate a group of Campylobacter-like organisms isolated from homosexual men suffering from enteritis, proctitis, or proctocolitis and has subsequently been isolated in cases of meningitis, bacteremia, and enteritis in humans, mainly those with immature or compromised immune systems (24, 31, 32). Recently, Weir et al. (35) described another helicobacter isolated from the blood of a patient with AIDS. All available information suggests that this isolate is H. cinaedi too. The whole-cell fatty acid and biochemical profiles conform with those of typical H. cinaedi strains (note that Trivett-Moore et al. [30] reported the "H. westmeadii" strains to be indistinguishable from H. cinaedi when examined by means of cellular fatty acid analysis). Indeed, the authors reported the absence of nitrate reduction as a criterion to separate their strain from H. cinaedi, but of a total of 26 H. cinaedi strains present in our biochemical database (data from the present study and from reference 2), 4 did not reduce nitrate. Absence of nitrate reduction is clearly not uncommon in H. cinaedi, and as discussed above, from the perspective of species level identification, the reliability of using 16S rDNA sequence similarity levels between 98.7 and 99.2% (35) is suspect, since intraspecies variation greater than this has been reported on more than one occasion. Our growing knowledge on the multiple reservoirs of H. cinaedi supports the hypothesis of Weir and coworkers that H. cinaedi (including "H. westmeadii" and Helicobacter sp. strain Mainz) may represent a Helicobacter species that is prone to cause sepsis in immunocompromised patients, such as those with AIDS. We now know that this bacterium occurs in hamsters (8), cats and dogs (16) (Table 1), and foxes and rats (Table 1). Its role in these animal hosts is not known. Good incubation and identification strategies will be required to establish the genuine significance of this Helicobacter species in bacteremic and enteric disease. Identification strategies by means of whole-cell protein or fatty acid analysis, extended biochemical testing, or restriction profile analysis of PCR amplicons derived from the 23S rRNA gene (12, 17) should be considered to support tentative identification results obtained by comparison of complete 16S rRNA genes. As useful as the latter method is, present data clearly indicate that it cannot be regarded as the "gold standard" for species-level identification of many members of the epsilon subdivision of the division Proteobacteria (Helicobacter, Campylobacter, Arcobacter, and related bacteria).