The French surveillance network of human Campylobacter infections is composed of laboratories selected on a voluntary basis (12). The laboratories are randomly located throughout France; 93 are hospital laboratories, and 338 are private laboratories. They send their clinical Campylobacter isolates to the National Reference Center with a completed information sheet concerning the patient and epidemiologic data. The study period was from July 2002 to December 2003, and 2,855 Campylobacter-like strains were studied. In addition to standard phenotypic identification, specific polymerase chain reaction (PCR) assays were carried out to identify C. jejuni, C. coli, or C. fetus (13); other isolates were identified by comparing 2 sequences of 400 bp located at both extremities of a 1,100-bp fragment of the beginning of the 16S rRNA gene amplified with primers F2-16S-CHPEC (ATCCTGGCTCAGAGTGAACG) and R2-16S-CHPEC (AAGGGCCATGATGACTTGAC) with those of DNA databases by using the BLAST program (http://www.ncbi.nlm.nih.gov/blast/). An identification at the species level was regarded as correct when >99% identity was seen with only 1 species. The species distribution from stool samples was as follows: 2,114 C. jejuni (78.9%), 486 C. coli (18.1%), 36 C. fetus (1.3%), 27 A. butzleri (1.0%), 8 C. lari (0.3%), 4 C. upsaliensis (0.1%), 2 C. hyointestinalis, and 1 Helicobacter canadensis. From 177 nonstool samples, we detected 2 A. butzleri.

Almost the entire 16S rRNA gene, 1,450 bp, was sequenced for 9 A. butzleri to study its homogeneity because only 5 sequences are in the databank (accession nos. AF314538, AY621116, L14626, U34386, U34388). The intraspecies variability was only 0.4% (6/1,430 bp) among the 14 A. butzleri sequences. Comparison of these sequences with those of 4 A. cryaerophilus (AY314755, L14624, U34387, U25805), 1 A. skirrowii (L14625), and 1 A. nitrofigilis (L14627) showed that interspecies variability concerned 14 bp out of 1,430. The Arcobacter sequences differed considerably from those of Campylobacter spp., with only 86.2% identity with C. coli (AY621115) and 85.6% identity with C. fetus (AY621110), 2 species that can be phenotypically confounded with A. butzleri.

Twenty-seven of the 29 A. butzleri strains were isolated from feces of patients with gastroenteritis, whereas 1 strain was isolated from peritonitis pus and another from a blood culture specimen. The 27 Arcobacter strains from stools were isolated in 23 different laboratories. Most of them (n = 15) used Campylosel (bioMérieux, Marcy l'Etoile, France) incubated at 37°C for 2 to 3 days. Six used Karmali agar incubated at 37°C for 3 to 4 days. The last 2 used Butzler medium and the filtration method, respectively. None of these infections was part of an outbreak. Of the 16 patients with travel information available, only 1 indicated a recent trip, to Turkey. An associated pathogenic organism was described in 1 of 15 cases, but the nature of the associated organism is unknown. Seventeen of the 29 patients were hospitalized. Their age range was 1–89 years, with an average of 54 years. Seventeen patients (59%) were male. Detailed clinical information was obtained for 19 patients and is shown in the Table. Eighteen of 19 patients had diarrhea, including the patient with bacteremia; 11 of 15 had abdominal pain; 4 of 15 had bloody stools, including the patient with A. butzleri bacteremia; 2 of 14 patients vomited; and 5 of 17 patients had fever. Acute renal failure developed in 2 patients, 1 associated with pyelonephritis. Concerning the 17 other cases for which information was available, severe clinical symptoms (anorexia, weight loss, asthenia) were described for 10 patients.

The duration of symptoms without treatment was variable, from 2 days to several weeks. Antimicrobial drug therapy was administered in 5 cases, and infection was eradicated a few days later (Table). In all cases, the strains were susceptible to the antimicrobial agent used (amoxicillin, ofloxacin, ciprofloxacin [twice], amoxicillin-clavulanate).

During an 18-month period, 2,855 strains of Campylobacter-like organisms were identified at the species level by using phenotypic and molecular tools. Molecular methods permitted the identification of 29 A. butzleri, i.e., 1% of the strains studied and also 1% of the 2,678 isolates from stools, which makes this species the fourth most frequently isolated Campylobacteraceae in human clinical samples in France, before C. lari and C. upsaliensis but after C. jejuni, C. coli, and C. fetus. Our results agree with those of Vandenberg et al. (11) who also found A. butzleri in fourth position for Campylobacter-like isolates from stool specimens. However, their results were obtained in a different context: only 1 laboratory was involved for 8 years, and a specific Arcobacter-selective medium was used. A study conducted in Denmark by Engberg et al. (14), also to estimate the prevalence of Campylobacter spp. and related bacteria in human fecal samples, did not produce the same results; only 1 A. butzleri was isolated out of 1,376 samples, and no C. lari or C. upsaliensis was recovered. The differences between these 3 studies could be explained by differences in geographic distribution of the species, but the difference in isolation methods (selective agar in the Belgian study) or in identification methods (molecular methods in our study) is likely the critical point. While not optimal, Campylobacter-selective agars, especially Campylosel, appear to allow the growth of A. butzleri. For the other media, a longer incubation period seems necessary. The role of A. butzleri in enteric infections is not definitively proven, but its involvement is likely. In at least 14 of 15 cases, no other enteropathogen was detected. In 1 case, a bacteremia with an enteric infection was found. These data are similar to those observed with C. jejuni infection. Isolation of A. butzleri is more frequent than well-known enteropathogenic Campylobacter spp., e.g., C. upsaliensis and C. lari, when adequate identification is carried out. The prevalence of A. butzleri may be underestimated because of false identification as C. coli or C. fetus. Indeed nearly all of the strains are able to grow at 25°C, like C. fetus, but are resistant to cephalothin like C. coli. We found that nearly all of the strains were resistant to nalidixic acid (4 had an intermediary resistance) and susceptible to ciprofloxacin with the disk diffusion method. The choice of a simple method for testing these drugs allows it to be used routinely in clinical laboratories to differentiate such strains phenotypically.

Nevertheless, molecular biology is a powerful tool for diagnosis. We used a 16S rRNA gene sequencing as a global approach for strains that were not C. jejuni, C. coli, or C. fetus. We confirmed that the intraspecies variability of A. butzleri 16S rRNA gene sequences is low (0.4%), lower than that found for C. coli (1.5%) or C. lari (2.5%) (15), and that interspecies variability is high. This approach, which is expensive and difficult to implement, is not accessible to a routine laboratory, but simple phenotypic tests can be applied to detect these microorganisms. The combination of phenotypic tests and selective agar medium will determine the real incidence of A. butzleri infection, which will add evidence that A. butzleri is an etiologic agent of diarrhea. A determination of the prevalence of A. butzleri in normal stool will also help establish the prevalence of this organism in diarrheal stool.