INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The bacterial genus Bartonella is a recently restructured taxon (3) which contains 14 species, among which 7 have been identified as human pathogens (12, 13, 22). These bacteria have been implicated in a large spectrum of infections. Bartonella quintana is the etiologic agent of trench fever, endocarditis, bacillary angiomatosis (22), and chronic bacteremia in homeless patients (4). Bartonella henselae can also cause both bacillary angiomatosis (17, 32) and endocarditis (8, 11) but is most commonly associated with cat scratch disease (7, 28) an illness which has also recently been attributed to Bartonella clarridgeiae (18). Only one isolate of Bartonella elizabethae has been obtained from the heart valve of a patient with endocarditis (5). Finally, Bartonella bacilliformis is responsible for bartonellosis (15).

Among the most profound clinical manifestation of Bartonella infection is endocarditis, a disease which often requires a surgical intervention with heart valve replacement and which is mainly caused by either B. quintana or B. henselae (22, 23). Early diagnosis of the infectious agent is important. Presently, serological techniques are most widely used, but shortcomings linked to cross-reactions with Chlamydia species (9, 19, 21) and variable sensitivities (22) have been reported. Histological examination of organ biopsy specimens is useful for the diagnosis of bacillary angiomatosis and peliosis hepatis, for example, but is not suitable for other clinical manifestations of Bartonella infections (22). Finally, biochemical procedures, such as cell wall fatty acid analysis, failed to dicriminate Bartonella spp. (5, 9, 33). Diagnosis can also be achieved by restriction fragment length polymorphism (RFLP) analysis of amplified DNA fragments such as the 16S-23S intergenic spacer region (ITS) (20, 30) or the citrate synthase gene (14, 26). Comparison of PCR-amplified genomic fragment sequences also leads to molecular identification of different bacterial species. Of these sequences, that of the 16S rRNA-encoding gene is by far the most widely used (35), and this approach has now become a standard method for the detection of pathogens (6, 34). However, the 16S rRNA genes of Bartonella species share more than 97.8% similarity (2, 3, 5). Thus, differences between them are not sufficient for confident discrimination of species (10, 31). Identification of Bartonella can, however, be reliably achieved by sequencing, for example, a portion of the citrate synthase gene (27). Indeed, DNA similarities of this 379-bp fragment are approximately 80 to 90% (9, 14). However, while PCR-based protocols offer several advantages over standard culture techniques, the risk of cross-contamination represents a major drawback. This is particularly true when numerous samples are manipulated in parallel. As a consequence, the finding of new PCR primers must be considered.

In the present study we assessed the usefulness of RNA polymerase beta-subunit-encoding gene (rpoB) sequence comparison as an alternative tool for identification of Bartonella spp. Comparison of rpoB sequences has been used for phylogenetic analyses among some members of the domains Archae (16, 25) and Bacteria (29). Recent work clearly illustrates that rpoB sequence analysis is a powerful tool for the identification of members of the family Enterobacteriaceae (24). In order to investigate the possible characterization of the genus Bartonella through rpoB gene sequencing, a 825-bp portion of this gene from 13 distinct strains was determined and further analyzed. A new PCR-RFLP method of Bartonella identification proposed from the present work was validated by analysis of clinical samples previously characterized by PCR sequencing of the ITS or the 16S rRNA gene.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and DNA sequencing. By using a QIAamp tissue kit (Qiagen, Hilden, Germany), DNA was purified from the Bartonella strains listed in Table 1 and from other bacterial strains used in this study, which included Chlamydia trachomatis, Chlamydia pneumoniae, Borrelia burgdorferi, Leptospira interrogans, Treponema pallidum, Serpulina pibscicot, Rickettsia prowazekii, Rickettsia rhipicephali, Staphylococcus aureus, Staphylococcus haemolyticus, Streptococcus sp., Francisela tularensis, the agent of human granulocytic ehrlichiosis, Listeria ivanovii, Campylobacter jejuni, Corynebacterium jeikeium, Escherichia coli, Pseudomonas aeruginosa, Legionella pneumophila, Mycobacterium tuberculosis, and Coxiella burnetii. PCR amplifications of rpoB fragments were performed with primers 1400F and 2300R (Table 2). These oligonucleotide sequences were deduced from a partial rpoB gene sequence of B. quintana obtained in the laboratory (unpublished data). For the reported experiments all primers used were either synthesized in the laboratory (392 DNA/RNA Synthesizer; Perkin-Elmer, Warrington, United Kingdom) or purchased from Eurogentec (Seraing, Belgium). Following a first denaturation step (94°C for 2 min), a three-step cycle of 94°C for 30 s, 53°C for 30 s, and 72°C for 1 min was repeated 35 times. The PCR program was ended by a single 2-min extension step at 72°C (Peltier thermal cycler model PTC 200; MJ Research Inc., Watertown, Mass.). Amplicons were then resolved by 1% agarose gel electrophoresis and visualized by staining with ethidium bromide. The QIAquick PCR purification kit (Qiagen) was used to prepare amplicons for automated sequencing, which was performed on an Applied Biosystems model ABI 310 automatic DNA sequencer (Perkin-Elmer) with dRhodamine Terminator Cycle Sequencing Ready Reaction Buffer (DNA sequencing kit; Perkin-Elmer) and primers 1400F and 2300R. In order to sequence the extremities of the 1400F-2300R region, two other primers, deduced from newly obtained sequence data, were used. The sequences of these oligonucleotides, designated 1596R and 2028F, respectively, are indicated in Table 2.

Clinical samples. Lymph node or pus aspirate samples from patients suspected of having cat scratch disease are sent to the Unité des Rickettsies for diagnosis. Detection of Bartonella DNA in these specimens was carried out by a previously described procedure (20). Briefly, a PCR was performed with primer pair QVE1 (TTCAGATGATGATCCCAAGC) and QVE3 (AACATGTCTGAATATATCTTC), which amplified a fragment of the 16S-23S rRNA intergenic spacer region. The amplified fragments were then sequenced (ABI 310 automatic DNA sequencer; Perkin-Elmer), and the resulting nucleotide alignments obtained were compared with those for sequences in both the public domain (GenBank) and our own laboratory database for final identification. Of the 94 samples received in the last 12 months, PCR amplification of the rpoB gene was also performed by using the conditions described above with primers 1400D and 2300R. The 21 positive amplicons, all previously identified as B. henselae, were then digested with ApoI (50°C overnight in the presence of bovine serum albumin). Finally, the DNAs from some of these samples were also sequenced.

PCR-RFLP analysis of Bartonella amplification products. Enzymatic digestion was performed by incubation of 5 µl of the purified PCR products obtained with primers 1400F and 2300R with appropriate buffer and 10 U of endonuclease. Following overnight incubation at the optimal temperature recommended by the manufacturer, the digestion products were separated on 1% agarose gels.

Data analysis. The nucleotide sequences of the rpoB gene fragments obtained were compiled and analyzed by computer with the autoassembler program of the ABI PRISM 310 Genetic Analyzer (Perkin-Elmer). Comparison of the sequences was performed with PC gene programs (Intelligenetics) by using the NALIGN program, and endonuclease sites were identified by using an in-house program (J. Gouvernet, unpublished data).

Nucleotide sequence accession number. The accession numbers of the sequences submitted to GenBank are given in Table 1.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Specific amplification of the Bartonella rpoB gene. By using primers 1400F and 2300R, an amplification product of 825 bp was obtained for all Bartonella strains analyzed. The PCR carried out with this pair of oligonucleotide primers was shown to be highly specific. Indeed, among the 21 other bacterial strains tested, no amplification products were observed. In contrast, and as expected, a positive response was obtained for all these strains by using 16S rRNA primers (data not shown).

Sequencing of the Bartonella rpoB amplified fragment. All PCR fragments were then sequenced at least in duplicate. The lengths of the fragments sequenced were always 825 bp, and neither insertions nor deletions were observed among the species analyzed (Fig. 1). The percentages of DNA similarity between pairs of strains are presented in Fig. 2. The comparison of the nucleotide alignments of the rpoB gene fragments from the Bartonella strains sequenced revealed levels of similarity between 84.9 and 99.8%. Two serotypes of B. henselae, namely, Houston and Marseille (8), exhibited a strong homology (99.8%), with only 2 bases among the 825 bases sequenced being different.

View larger version (112K): [in this window] [in a new window]   FIG. 1.   Alignment of nucleotide sequences of the Bartonella rpoB genes amplified by PCR. The primers used were 1400F and 2300R. Homologies are indicated by dots. 1, B. alsatica; 2, B. bacilliformis; 3, B. berkhoffii; 4, B. clarridgeiae; 5, B. doshiae; 6, B. elizabethae; 7, B. grahamii; 8, B. henselae Houston; 9, B. henselae Marseille; 10, B. quintana; 11, B. taylorii; 12, B. tribocorum; 13, B. vinsonii.

View larger version (21K): [in this window] [in a new window]   FIG. 2.   Similarity values between rpoB sequences of various Bartonella species. Values were deduced from the data presented in Fig. 1 by comparison of 825 nucleotides. BhensH, B. henselae Houston; BhensM, B. henselae Marseille; the remaining abbreviations across the top correspond to the species on the left, from top to bottom, respectively.

PCR-RFLP identification of Bartonella. The suitabilities of a large number of endonucleases were then assessed for all the Bartonella strains sequenced by using a homemade program (Gouvernet, unpublished data). From the resulting analysis it appeared that the combination of both successive digestions allowed easy discrimination of these strains. Thus, ApoI digestion led to four different patterns. B. quintana was the only strain for which four predicted fragments were obtained, thus allowing its direct identification and its placement in the first group. At the opposite extreme, neither B. bacilliformis nor Bartonella grahamii was shown to have an ApoI restriction site, and both species were classified in the second group. In the third group, we found B. henselae (serotypes Marseille and Houston) as well as Bartonella alsatica, for which the pattern was two bands of 735 and 87 bp, respectively. Finally, the seven other strains were all hydrolyzed near the middle of the inititial 825-bp PCR fragment, leading to two bands of approximately 400 bp each. In fact, two subgroups were obtained. In the first subgroup, the ApoI site was located at position 434 (B. elizabethae and Bartonella tribochorum) and in the second subgroup it was located at position 471 (Bartonella berkhoffii, B. clarridgeiae, Bartonella doshiae, Bartonella taylorii, and Bartonella vinsonii). However, the RFLP patterns of both subgroups were too close to permit their differentiation in agarose gels. This analysis was validated by the experimental data presented in Fig. 3 and 4. Under our experimental conditions, the small 37-bp fragment expected for the B. quintana strain was not detected, probably because of the sensitivity of the method. However, the profiles obtained allowed differentiation of the four patterns. Moreover, and as illustrated in Fig. 4, the digestion profiles obtained with ApoI allow easy differentiation of B. henselae and B. quintana strains. This is of importance when considering the fact that both of these bacterial species are implicated in endocarditis, which is the most common clinical manisfestation of Bartonella infection (13). By using a second endonuclease, all the Bartonella strains from the four previously determined groups can be definitively identified. Such a differentiation can be reached with the restriction enzymes listed in Tables 3 to 5, among which were included AluI and AflIII. Indeed, AluI digestion of the rpoB amplicons of the group 2 Bartonella yielded either five or three fragments, and thus allowed easy differentiation of B. bacilliformis and B. grahamii, respectively. This enzyme was also shown to be efficient in the differentiation of the bacteria included in group 4 since a specific profile was obtained for each of the seven strains included in this group. Finally, the differentiation of B. alsatica and both serotypes of B. henselae (group 3) can be achieved by using AflIII. While the B. alsatica rpoB fragment was devoid of such a restriction site and consequently was not hydrolyzed by AflIII, it was not the case concerning B. henselae, which was digested into three and four fragments for serotypes Houston-1 and Marseille, respectively. In fact, when combined with a first digestion step with ApoI, all of the restriction enzymes presented in Tables 3 to 5 allowed identification of all the Bartonella species by RFLP analysis.

View larger version (71K): [in this window] [in a new window]   FIG. 3.   Restriction profiles obtained after digestion of a portion of the rpoB gene with ApoI. Ethidium bromide-stained agarose gels of ApoI restriction endonuclease digests of DNA amplified by using primers 1400F and 2300R are shown. Lane A, molecular mass markers (marker IV; Boehringer). Ba, B. alsatica; Bba, B. bacilliformis; Bbe, B. berkhoffii; Bc, B; clarridgeiae; Bd, B. doshiae; Be, B. elizabethae; Bg, B. grahamii; Bh1, B. henselae Houston; Bh2, B. henselae Marseille; Bq, B. quintana; Bta, B. taylorii; Btr, B. tribocorum; Bv, B. vinsonii. Numbers on the left are in base pairs.

View larger version (72K): [in this window] [in a new window]   FIG. 4.   ApoI digestion profiles of rpoB amplicons from either B. henselae or B. quintana. Ethidium bromide-stained agarose gels of ApoI restriction endonuclease digests of DNA amplified by using primers 1400F and 2300R are shown. Lane A, molecular mass (marker IV; Boehringer); lanes 1 to 5, DNA extracts from blood of patients infected with B. quintana; lanes 6 to 10, DNA extracts from lymph node or pus aspirate samples from patients suspected of having cat scratch disease and identified as B. henselae-positive samples. Numbers on the left are in base pairs.

Confirmation of diagnosis obtained with clinical samples. Among the 94 samples tested, 21 were found to be positive for Bartonella. This result was deduced from PCR assays performed both with ITS primers used under established conditions and with rpoB primers specific for Bartonella and designed in this study, namely, primers 1400D and 2300R. Thus, each positive or negative result obtained by the ITS PCR methodology was confirmed by rpoB PCR (Table 6). All amplification products were identified by sequencing of the ITS amplicon as being derived from strains of B. henselae. This result was confirmed by digestion of all fragments obtained with ApoI and by sequencing of some of the rpoB amplicons. During the same period, 24 isolates were collected in the laboratory from the blood of either patients or cats. PCR sequencing of ITS identified these strains as follows: 10 B. quintana isolates, 4 B. clarridgeiae isolates, and 10 B. henselae isolates. All these isolates were also analyzed by PCR-RFLP analysis of the rpoB gene. In all cases, the patterns of digestion obtained with ApoI and AluI corroborated the diagnosis previously deduced from ITS sequence analysis.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Several techniques for detection of Bartonella species in clinical material have been described elsewhere (23). Among these, PCR-based protocols present several advantages over culture. For example, they allow the detection of slowly growing pathogens or bacteria in fixed biopsy material. Nevertheless, some limitations must be considered, in particular, the potential for contamination of samples to lead to false-positive results. This may be remedied in part by use of different rooms for procedures upstream and downstream of the amplification step. Currently, the 16S rRNA gene has been the focus for most PCR methods because it is one of the most conserved genes (32, 34, 35). However, for the identification of Bartonella, this approach is not considered satisfactory due to the high percent similarity of the sequence of this gene among different Bartonella species (10, 31). Conversly, identification of these bacteria can be more reliably achieved by either ITS (20, 30) or citrate synthase gene (9, 14, 27) PCR assays. However, because of the potential risk for contamination discussed above, there remains a need for the development of alternative assays. We thus attempted to develop such an assay based on the rpoB gene. This gene has previously been used for phylogenetic analysis among some members of the domains Archae (16, 25) and Bacteria (29) and has been demonstrated to be a powerful tool for the identification of members of the family Enterobacteriaceae (24).

Initially, different pairs of primers used to amplify the rpoB gene fragments of members of the family Enterobacteriaceae were tested with Bartonella. From the preliminary sequences obtained, two primers which were observed to be specific for Bartonella species were deduced. The specificities of these primers, designated primers 1400F and 2300R, were then demonstrated by a PCR assay involving 21 bacterial strains unrelated to members of the family Bartonellaceae. In a second step, the rpoB gene fragments of 13 distinct species were amplified and automatically sequenced. Analysis of the resulting nucleotide sequences demonstrated that ApoI digestion of the initial 825-bp PCR fragment allowed differentiation of B. quintana from all other species. This point is of importance when one considers the importance of this strain in human infections (22). The other species each yielded one of three ApoI profiles but could be distinguished from one another if those profiles were combined with the results of a second digestion with either AluI (group 2 and 4) or AflIII (group 3). This rpoB sequencing methodology was validated from experiments performed with DNAs from clinical isolates as the DNA template. Indeed, all B. henselae-positive samples identified by routine methodology from lymph nodes or pus aspirates were, without exception, shown to be positive by the rpoB PCR done with our Bartonella-specific primers. These data were validated by the digestion of all the amplicons obtained with ApoI, for which the pattern was two bands of 735 and 87 bp, respectively. As indicated in the Results section, the same profile of digestion can be achieved with B. alsatica. However, this bacterial species has been isolated from the blood of rabbits and has never been demonstrated to be responsible for infection in humans (13). While there was no doubt, the initial diagnosis was firmly confirmed for some samples by the sequencing of the rpoB amplicons. Similarly, the identifications obtained by PCR-RFLP analysis with clinical or animal isolates were also in accordance with previous identifications. Finally, alignment of Bartonella rpoB sequences allowed the design of a new internal primer chosen from the most conserved regions of the rpoB gene region sequenced (1873R, underlined part of the sequence in Fig. 1; Table 2). When used in combination with primer 1400F, a 439-bp PCR product was obtained (data not shown). Sequencing of such an amplified portion of the gene could be an alternative for identification of these bacteria in properly equipped molecular biology laboratories.