The bacteria isolated from the patient grew slowly on the chocolate, Haemophilus chocolate, and BCP agar plates. On routine media, the bacteria were irregular gram negative coccobacillus, nonmotile, and strictly aerobic. They tested positive for oxidase and catalase, were susceptible to colistin, and grew at 36°C, 25°C, and 42°C.

Automated routine identification showed twice that this bacterium was in the Moraxella group (probability 95%). API 32GN identified the bacteria at 24 h as Methylobacterium mesophilicum (doubtful identification: 98.4%, with low typicity index [T] = 0.39) and at 48 h as Acinetobacter, Pseudomonas, or Achromobacter (nonreliable identification). API 20NE identified the bacteria (profile code 1001067) with weak discrimination as Achromobacter denitrificans (31.1%; T = 0.59) or B. bronchiseptica (67.1%; T = 0.66). RapID NH gave Haemophilus influenzae as first choice (inadequate identification; probability 76%; bioscore 1/610).

At 48 h to 72 h, cultures on BGA medium showed small, nonhemolytic colonies ≈1 mm in diameter, which were not producing any brown pigment. These tested negative for urease production. The spectra obtained gave a good identification of B. petrii (identification agreement 55.4%).

The sequencing of the 16S rRNA confirmed the MALDI-TOF identification. The 16S sequences showed 99% of similarity with the 16S rRNA gene sequences from the type strain of environmental origin and the other isolate from human origin. We then performed the sequencing of risA and ompA genes to compare with the genes of the other B. petrii isolates. The risA gene sequences of the clinical isolates showed 93% similarity with the type strain of environmental origin. The species with the next highest similarities were B. bronchiseptica, B. parapapertussis, and B. pertussis, all with a nucleotide identity of 87%. The ompA gene sequences of the clinical isolates demonstrated 89% similarity with the type strain of environmental origin. The species with the next highest similarities were B. bronchiseptica, B. parapapertussis, and B. pertussis, all with a nucleotide identity of 86%. The weaker score obtained with ompA was due to an insertion of 12 nt versus the environmental strain, nucleotides not present in the other isolates from human origin.

The data regarding antimicrobial drug susceptibility obtained are shown in Table 2. Results are compared with other data available from the literature.

As shown in Figure 1, the PFGE patterns obtained with the DNA from the 3 isolates of the reported human case-patient are identical, whereas the patterns obtained with the DNA of the other isolate from human origin and the isolate from the environment show several differences. This indicates that the isolates of the present study are related but part of a different PFGE group.

Results were negative for Fim2 and Fim3 by using the agglutination technique as were results for FHA, PRN, PT, and AC-Hly by Western blot (data not shown) with specific antibodies. As shown in Figure 2, the serum of the patient infected with B. petrii recognized several proteins in the bacterial suspensions of the 3 B. petrii isolates collected from the patient. The same proteins are recognized in the bacterial suspensions of the 2 other B. petrii isolates of clinical and environmental origin, except for 1 low-molecular-weight protein. Most of these proteins, with small differences in molecular weights, are also recognized in bacterial suspensions of B. pertussis, B. parapertussis, B. bronchiseptica, and B. holmesii, except for 1 protein. The pools of serum samples from patients infected with B. pertussis and B. bronchiseptica recognized high-molecular-weight proteins expressed by B. pertussis, B. parapertussis, and B. bronchiseptica in the bacterial suspensions of these 3 species. However, they did not recognize these high-molecular-weight proteins in the bacterial suspensions of B. petrii and B. holmesii. Finally, the pool of serum samples from patients infected with B. bronchiseptica recognized 1 protein in the B. pertussis, B. parapertussis, B. bronchiseptica and B. holmesii bacterial suspensions but not in the B. petrii bacterial suspensions.

In terms of cytoxicity, B. pertussis and B. bronchiseptica are cytotoxic for the J774-A1 macrophages. However, none of the B. petrii isolates were cytotoxic.

Identification of B. petrii is still a major problem for clinical laboratories that use automated or manual identification systems. As suggested by Zbinden et al. (37) isolates that do not give a 99% or better typing result should be typed by 16S rRNA sequencing or MALDI-TOF. The spectra obtained here with MALDI-TOF gave an acceptable identification of B. petrii (identification agreement 55.4%). This is a good score, especially because the database contains only 5 spectra of this recently described species because of the low number of isolates available. The phenotypic characteristics of the isolates in our study are similar to those of the few isolates that have been previously described (1,16–19).

In a previous study on B. bronchiseptica, we and others working on Bordetella spp (16). determined that the results obtained for many antimicrobial drugs using the disk diffusion method correlated poorly with clinical therapeutic results and with MICs established using the reference method ( 32,33; A. Le Coustumier, unpub. data). Fry et al. (16) reported that the clinical isolate was apparently susceptible, by disk diffusion tests, to 5 antimicrobial drugs: clarithromycin, erythromycin, gentamicin, ceftriaxone, and piperacillin+tazobactam. However, the respective reference MICs indicated that only piperacillin+tazobactam was active in vitro with a MIC of 2 µg/mL. Based on the preliminary results the patient received a 6-week course of oral clarithomycin treatment. Despite the successful clinical outcome, the isolate was subsequently shown to be resistant to clarithromycin in vitro. In the only other report (to our knowledge) on a clinical B. petrii isolate, MICs were determined by using VITEK2 Compact (bioMérieux) but MICs of drugs for Bordetella spp. cannot be determined from this database (18). Using Etest strips, a method that has been validated on a wide range of glucose fermenting and nonfermenting gram-negative bacteria, we determined the MIC for 26 widely used antimicrobial drugs from the main therapeutic families (38).

All of the 5 isolates in the present study as well as the isolates described by Fry et al. (16) and Stark et al. (18) appear to have resistance to penicillins (penicillin, amoxicillin), cephalosporins (especially third-generation, extended-spectrum cefotaxime or ceftriaxone and ceftazidime), clindamycin, quinupristin and dalfopristin, rifampin, linezolid, daptomycin, and fucidic acid. We also observed that aminoglycosides had only moderate activity against the bacteria. The isolates in our study also displayed in vitro sensitivity, but low level MICs, to minocyclin, tygecyclin, cotrimoxazole, and fosfomycin.

The large gap in the MICs of amoxicillin and piperacillin between the only environmental isolate available for this study and the clinical isolates may reflect the inducible response to exposure of the clinical isolates to formerly widely used treatment with β-lactams. Tazobactam does not restore the activity of piperacillin, or even degrade it, probably because of the induction of β-lactamase.

In contrast with the isolate of Fry et al. (16) and the environmental isolate (1), the MICS of carbapenems and systemic fluoroquinolones were high for the isolates from our patient. No previous treatment with carbapenems could be documented from the long medical history of our patient, although fluoroquinolones had been frequently prescribed for bronchiectasis. This lack could be partly due to an impermeability-linked cross-resistance between these 2 chemically unrelated families with the common porin mutation, as frequently has been observed for Pseudomonas aeruginosa.

Using PFGE, we observed that the patterns of the DNA restriction fragments for the different isolates collected from the reported human patient were quite similar. This finding confirms the persistence of the same isolate inside the host. However, several differences are observed with the patterns of the DNA from the environmental or human isolates. These differences could be linked to the loss of pathogenic islands in some of the isolates, as has been recently reported (22).

Using murine serum samples specific to the major virulence factors expressed by B. pertussis and pool of sera from patients infected with either B. pertussis, B. bronchiseptica, B. holmesii, or the serum of the current patient infected with B. petrii, we confirmed that B. petrii isolates do not express FHA, Fim2 and Fim3, PRN, PT, and AC-Hly. The serum sample from the patient infected with B. petrii recognized only 1 protein specific to the B. petrii bacterial suspensions derived from the isolates of the clinical patient described in this study. Another protein was specific to the 5 B. petrii isolates.

None of the B. petrii isolates were cytotoxic for macrophages. This result was likely because these isolates do not express AC-Hly or BteA.

The source of infection and the pathogenic role of B. petrii are still unknown. For the study patient, the source of infection, just prior to the first episode, was most likely a contamination that occurred during the aerosol therapy performed at home under poor hygienic conditions (according to the patient). This was potentiated by local corticotherapy.

The prevalence of Bordetella spp. within the cystic fibrosis population may well be underestimated, due to the slow growth of this microorganism. However, the prevalence may also be underestimated for all immunosuppressed patients, particularly the elderly. The role that Bordetellae spp. such as bronchiseptica and petrii may play in the progression of pulmonary disease remains unknown, and these species can be misidentified in hospital laboratories (19).