Clinical and epidemiological analysis of Campylobacter fetus subsp. fetus infections in humans and comparative genetic analysis with strains isolated from cattle

Bacterial CFF isolates from 17 patients were identified between 2001 and 2009 at our institution, a university hospital in Switzerland with specialized microbiological laboratories serving an estimated population of 1.2 million. Confirmation of human CFF infections was based on positive culture results. C. fetus was identified by sequencing 500 bp of the 16S rRNA gene. Clinically relevant patient information was obtained from medical charts and supplemented by telephone interviews with patients and treating physicians. Seventeen bovine isolates originated from faecal samples of healthy calves from different geographic regions of Switzerland. Bacterial strains were cultured on tryptone soya agar containing sheep blood 5 % v⁄v (Oxoid, Wesel, Germany) under microaerobic conditions for 2 days at 37 °C. C. fetus subsp. identification was based on glycine tolerance test. The results were confirmed by PCR [21].

PFGE was performed using the restriction enzyme SmaI as previously described [21, 22]. The DNA banding pattern was analyzed using BioNumerics software (version 7.5; Applied Math, Sint-Martens-Latem, Belgium) and a dendrogram was generated by cluster analysis using a similarity matrix (settings: Dice, UPGMA with 3 % optimisation and 1 % band tolerance). A cut-off at 70 % similarity of the Dice coefficient was applied to indicate major PFGE clusters. Fisher’s exact test was used for statistical analysis of the dendrogram.

The 2 MOMP encoding genes cmp1 and cmp2 in CFF were partially sequenced. Primers were designed based on the reference genome sequence of C fetus subsp. fetus 80–40 (GenBank acc. no. NC_008599) to amplify fragments encoding the C-terminal part of these proteins which contain putative external loop structures, and are predicted to be the most variable regions in these proteins. The 3′ part of the gene cmp1 was amplified with primer pair CFMOMP1_2-L (5′-TTGAAAATGTTACTGATCTATACGCTATAGAC) and CFMOMP1-R (5′-AGTTTTATCTTTACCAGCATTGTCTGGCTCTT CAGTT) resulting in a 536 bp PCR product (position 650–1185 of cmp1). The 3′ part of cmp2 was amplified with primer pair CFMOMP1_2-L and CFMOMP2-R (5′-GTTTTTGAAATCGTTTTTGCCATCTATC) resulting in a 545 bp PCR product (position 641–1185 of cmp2). Sequence alignments were performed using the Vector NTI advance software (Invitrogen), and a unique arbitrary number was given to every single allele based on the reference sequences of cmp1 and cmp2 of strain 80–40 defining allele number 1 (GenBank locus tag CFF8240_0396 and CFF8240_0397, respectively). The four allele sequences of cmp1 and the five of cmp2 have been deposited at GenBank under accession numbers KU670818-KU670821 and KU670822-KU670826, respectively. A full matrix including sequences of all strains analyzed in this study can be accessed at http://​purl.​org/​phylo/​treebase/​phylows/​study/​TB2:​S18853.

The MLST analysis was carried out using the protocols of the C. fetus MLST database as described before [14]. New alleles and MLST sequence types (ST) were deposited in the PubMLST database (http://​pubmlst.​org/​cfetus/​). PCR screening for the presence of the T4SS and susceptibility tests determining MIC for tetracycline and streptomycin were carried out as previously described [20]. The analysis of the sap/serotypes was done by PCR as published [23].

In humans, diagnosis of CFF infection was based on positive blood cultures in 76.5 % of cases, and on other sources in 41.1 % of cases (stool, abscess fluid, biopsies; in addition to blood cultures in cases 1, 9, 12; Table 1). CFF was mostly isolated from elderly patients (mean age, 65 years; range 31–87 years), with a slight preponderance of male patients (10 males, 7 females). The site of infection was the gastrointestinal tract (colitis) in 6 cases, the native venous tract in 4 cases (deep venous thrombosis in a leg, in 1 patient (case 11) associated with erysipelas), the skin (erysipelas) in 2 cases, the site of foreign material (3 joint prostheses, 1 vascular Y-prosthesis, 1 implanted central venous catheter Port-a-Cath®) in 5 cases, and in 1 patient the urinary tract. Abscess formation was noted in 3 cases: 2 psoas abscesses associated either with colitis (case 9) or with infection of the hip prosthesis (case 15), and 1 abscess in the deep tissues of the plantar foot associated with thrombosis in the leg (case 12). Of the 6 patients with a joint prosthesis, only 3 had the infection at the site of the implant. Seven patients were immunocompromised, confirming previous observations for susceptibility to CFF infection. Underlying diseases included most commonly heart disease (9 patients), arterial hypertension (8 patients), renal failure (5 patients) and type 2 diabetes mellitus (4 patients), underscoring the morbidity of this mostly elderly population. Interestingly, the presence of diverticulosis (of the sigmoid colon in patients 2, 13 and 15, and of the whole colon in patients 3, 14 and 17) did not predispose to gastrointestinal infection with CFF, as only patients 14 and 17 became symptomatic, whereas colitis was a presenting symptom in patients 1, 4, 6 and 9 who did not have known diverticulosis. All but patients 10, 11 and 12 were moderately anaemic (82 %, haemoglobin 90–110 g/L). We found two co-infections: in patient 1 with Clostridium difficile and in patient 9 with Campylobacter jejuni (both isolated from stool specimens). Possible bacterial entry routes were skin or mucosal lesions in at least 14 out of the 17 patients; for patient 2, this information was not available. Skin lesions were open wounds including a leg ulcer (patient 11), whilst mucosal lesions included open wounds after extraction of a wisdom tooth and removal of a mandibular cyst (patient 1), oral ulcers (patient 5), ulcerous cavities due to poor dental status (patient 9), erosive gastritis (patient 7) and oesophagitis (patient 4).

Most of the patients (59 %) were treated with quinolones (ciprofloxacin or ofloxacin), especially when gastrointestinal symptoms were present. In the case of colitis, clarithromycin was successfully used (patients 4 and 14). In the cases of soft tissue infections (erysipelas, abscess), either in the presence or absence of venous thrombosis, amoxicillin was more readily used. The use of antibiotics was always adapted to the susceptibility results of the isolates. In the patients with prosthetic device infection (joints and catheter), only complete removal of the synthetic material and concomitant, longstanding antibiotic therapy led to the resolution of the infection. For example in patient 15, symptomatic CFF infection recurred 6 years after the first CFF infection and antibiotic therapy; the patient was finally cured after removal of the osteosynthetic material. In patients 5 and 10, antibiotic therapy alone over years (2 and 3 years, respectively) led to control of the chronic infection and formation of a fistula.

At the time of CFF bacteremia, patient 13 had a PET/CT scan showing ¹⁸F-FDG uptake at the site of the aortoiliac Y-prosthesis, suggesting infection of the prosthesis which was implanted 2 years before. Antibiotic therapy alone (meropenem followed by oral doxycycline for 3 months) led to resolution of the inflammatory signs in the blood and negative blood cultures on follow-up. None of the 17 patients died due to CFF infection.

Epidemiological data reveal contact with animals for 59 % of the patients. However only a minority had contact with animals known to be carriers of CFF: horses (patient 5), cattle (patient 11 and 12), pigs (patients 12 and 13), sheep (patient 15) and turtles (patient 17). All but one subject (patient 8) either consumed raw or undercooked meat (patients 4, 5, 9, 10, 14), raw eggs (patients 2, 4, 7, 9, 13) and/or raw vegetables (patients 1, 2, 6, 7, 10, 11, 12, 13, 15, 16, 17). Patient 13 was a butcher, and patients 1, 2, 3, 5, 8, 9, 10, 11, 12, 13, 15, and 16 (71 %) were regularly visiting farms or were working there (patients 11, 12 and 15 were farmers). One patient (case 6) frequented a swimming pool, and only 2 patients (cases 10 and 15) had contact with young children. Traveling was limited: 9 patients never traveled, 5 patients visited other small localities within Switzerland occasionally, and 3 patients had been to Southern Germany, Spain, or South Africa in the last 3 years.

The locus tags CFF8240_0396, and CFF8240_0397 of the CFF strain 82–40 (GenBank acc. No: NC_008599) were used as a reference for cmp1 and cmp2, respectively. The nucleotide diversity of the 3′ part of both cmp genes was analyzed in all the isolates, and arbitrary numbers setting the reference sequence as 1 were used to distinguish the allele sequences (Table 2). Fifteen and 22 variable nucleotide positions were identified in cmp1 and cmp2, respectively. Silent substitutions are indicated in lower cases, non-silent ones in upper cases. They were calculated using the in frame predicted amino acid sequences of both genes. Interestingly, some of the nucleotide substitutions are shared between both, equally long cmp gene sequences, indicating most likely a gene duplication event and/or the same purifying selection. The allele type of both genes correlated highly with the clusters obtained by PFGE and MLST, and a similar level of discrimination was observed as for the PFGE and MLST analyses (Fig. 1). Moreover, cmp gene sequence types demonstrate a 100 % discrimination between cluster A and B. With cmp1 only type 2 was found in cluster A, i.e. in the human as well as the bovine isolates. In cluster B, only cmp1 sequence types 1, 3, and 4 were found. Type 3 was found in the bovine as well as in human isolates. With respect to cmp2, sequence types 2 and 5 were only found in cluster A, with 2 found in both, human and bovine, while 5 only in bovine isolates. Types 1, 3 and 4 were only detected in cluster B with type 3 in isolates from both hosts.

Additionally, the predicted amino acid sequences encoded by cmp1 and cmp2 of the reference CFF strain 82–40 were further analyzed. Using the TMB-Hunt program (http://​www.​bioinformatics.​leeds.​ac.​uk/​), we classified both proteins as transmembrane beta barrel proteins using whole sequence amino acid composition [24]. Therefore, the topology of the outer membrane proteins with respect to lipid bilayer was predicted with the PRED-TMBB web server (http://​biophysics.​biol.​uoa.​gr/​PRED-TMBB/​). In the cmp1 encoded protein, 16 transmembrane domains and 8 external loops were identified, whereas in the cmp2 encoded protein 20 transmembrane domains and 10 external loops were found. In both, none of the predicted amino substitutions affected the predicted topology of the proteins.

The T4SS was identified in 11 isolates of cluster A (55 %) and in 1 isolate of cluster B (7 %), in bovine as well as in human samples. There was an expected and complete concordance of the presence of T4SS and resistance to tetracycline and streptomycin, as indicated by higher MIC values. These results suggest that CFF strains with the T4SS all carry a transferable pathogenicity island. Complete genetic analysis and demonstration of the presence of the antibiotic resistance genes tet(44) and ant(6)-Ib has been shown elsewhere in detail for the bovine clone IMD_523/06 [20].