Cross-Infection Due to Imipenem-Resistant Bacteroides fragilis Associated with a Totally Implantable Venous Port

Abstract

Four patients in an oncology ward developed Bacteroides fragilis bacteremia over a 12-day period. Cross-infection between two of them, due to an imipenem-resistant strain, was demonstrated by epidemiological investigation and genotypic typing methods (arbitrarily primed PCR fingerprinting and nucleotide sequencing of the cfiA genes and upstream IS1186/IS1168 elements).

Bacteroides fragilis is often responsible for intra-abdominal sepsis and as such is an important cause of nosocomial infections (7-9, 12, 14, 17). However, very few studies have been conducted on the possible dissemination of this organism within hospitals, probably because it is typically an endogenous pathogen (6, 25). Earlier studies based on serotyping suggested that strains of B. fragilis might spread between hospitalized patients (4, 5). But more recently, by using molecular typing techniques, no clonal distribution could be detected within B. fragilis hospital isolates (13). We report here a case of cross-infection due to an imipenem-resistant B. fragilis strain, as demonstrated by epidemiological investigation and genotypic typing methods.

Over a 12-day period in September 2001, four patients hospitalized in the oncology ward of Saint-André Hospital in Bordeaux, France, presented B. fragilis bacteremia (Fig. 1) despite the low incidence of this organism among bloodstream isolates (ca. 3%) (14, 17). Of the four patients, only patients A (kidney carcinoma) and D (colic adenocarcinoma), both admitted for a subocclusive syndrome, were likely to develop B. fragilis bacteremia. Indeed, members of the B. fragilis group reside in the gut and, from this reservoir, can colonize the female genital tract but they are virtually absent in the upper respiratory tract (6, 12). Accordingly, B. fragilis bacteremia originates much more frequently from abdominal (60 to 70%) rather than pelvic (3 to 5%) processes (patient B, cervical cancer) and almost never from respiratory disease (patient C, bronchial epidermoid carcinoma) (7-9, 14, 17). Nevertheless, all four patients had risk factors for B. fragilis bacteremia that included malignancy and use of immunosuppressive therapy (14, 17).

FIG. 1.

Clinical cases. Abbreviations: M, male; F, female; y, years old; sd, syndrome; BC, blood culture; d, day; †, death.

No epidemiological link could be found between the four patients, except for the insertion of a totally implantable venous port (IVP) in patients B and C on the same day, in the same operating room. The timetable for the event in this room on this day included a gastro-jejunal derivation with cholecystectomy performed in the morning by a surgical team in patient X (7:35 to 12:20 a.m.) which ended with compressive dressing of the wound with Mefix tape (Molnlycke Health Care, Warenne, Belgium). After a period during which the room was cleaned and thereafter not entered by anyone (12:20 a.m. to 3:00 p.m.), three Port-A-Cath systems (Sims, Orly, France) were implanted in patient B (3:00 to 3:45 p.m.), an unrelated patient Y (4 to 5 p.m.), and patient C (5:15 to 6:00 p.m.) by a second surgical team. In contrast to patient Y, patients B and C had an IVP outfit with a Huber needle secured by the sticking Mefix bandage instead of an ordinary dressing and they subsequently developed an IVP-related infection, as defined by bacteremia without another apparent infection focus and/or inflammation or discharge at the insertion site (15). Infection is a rare complication of IVPs, and the predominant pathogens are staphylococci (10, 15).

The nonsterile bandage used to fix the IVP and the Huber needle was suspected to be the source of the contamination. Actually, adhesive tapes that secure Huber needles are applied in close contact with the intravascular insertion site for extended periods of time and can contribute to local infections (23). A sampling campaign (12 sites), carried out in the same operating room 1 month after the initial episode showed that several articles of noncleanable, nonsterilizable equipment, including the Mefix tape, were contaminated by skin-commensal and environmental organisms. The roll of tape may have been contaminated by the digestive flora of the patient operated on in the morning (patient X). Anaerobes substantially outnumber aerobic organisms in the digestive flora, and B. fragilis appears to be one of the most virulent (6, 7-9, 25). Moreover, the B. fragilis strain Bf1149 was demonstrated to remain viable on the adhesive tape for at least 8 h when present at levels higher than 10⁶ CFU/ml (data not shown), in agreement with the fact that B. fragilis is only a moderate obligate anaerobe (26).

The four strains of B. fragilis isolated from patients A to D gave identical or similar biochemical reactions (12) (Table 1). Antibiotyping by the disk diffusion method (23 tested antibiotics) showed that Bf1149 and Bf1150 were resistant to all β-lactams including imipenem, in contrast with the two other strains (MICs of 128 mg/liter and 0.5 mg/liter, respectively) (Table 1). Imipenem resistance is an exceptional trait among B. fragilis strains (0 to 3%) (6, 16, 21). This resistance is due to the production of a group 3a class B metallo-β-lactamase encoded by the silent cfiA chromosomal gene, the expression of which is promoted by the insertion of an insertion element (IS) immediately upstream (1, 18, 19, 21).

TABLE 1.

Characteristics of the B. fragilis strains used in this study

Strain	Patient	Hospital unit	Clinical specimen	Date (day/mo/yr)	Methoda		Drug resistance byb antibiotyping
					API 20 A code	Rapid ID32A code
Bfl148	A	1	Blood	11/09/00	46546240	4517453332	TIC, TET
Bfl149	B	1	Blood	14/09/00	46544240	4517453332	TIC, TIM, IPM, TET
Bfl150	C	1	Blood	14/09/00	46544240	4517453332	TIC, TIM, IPM, TET
Bfl151	D	2	Blood	23/09/00	46546240	4517453332	wtc
ATCC 25285					46544240	4517453332	wt

^abioMérieux, Marcy l'Etoile, France.

^bTIC: ticarcillin; TET: tetracyclines; TIM: ticarcillin-clavulanic acid; IPM: imipenem.

^cwt, wild-type susceptibility.

In order to type the four strains of B. fragilis by a molecular method, arbitrarily primed PCR fingerprinting with four primers was undertaken (2, 18). Bf1149 and Bf1150 gave strictly identical patterns, whereas Bf1148 and Bf1151 yielded different and distinct profiles (Fig. 2). Previous PCR fingerprinting experiments have demonstrated that the cfiA-positive B. fragilis strains yield a homogeneous DNA fragment pattern (2, 18). The same data were obtained with reference genotypic typing methods (18). Thus, conventional genotypic methods are inadequate for typing the genetically homogenous cfiA-positive B. fragilis.

FIG. 2.

Arbitrarily primed PCR fingerprinting of B. fragilis strains. Fragments were generated with AP12h (5′-CGGCCCCTGT-3′), AP2 (5′-ATTGCGTCCA-3′), ERIC1R (5′-ATGTAAGCTCCTGGGGATTCAC-3′), and ERIC2 (5′-AAGTAAGTGACTGGGGTGAGCG-3′). Lanes ATCC, 1148, 1149, 1150, and 1151 were patterns obtained from B. fragilis strains ATCC 25285, Bf1148, Bf1149, Bf1150, and Bf1151, respectively. Lanes M were size marker profiles (PstI fragments from the λ phage DNA).

The presence of the cfiA gene and an upstream IS in Bf1149 and Bf1150 was confirmed by PCR (27). Sequence analysis of the 2,250-bp amplified DNA fragment (GenBank sequence database accession number AF429432) showed that the 747-bp cfiA genes of Bf1149 and Bf1150 were identical and differed from the most closely related cfiA sequence available in GenBank by eight mismatches, including two missense mutations (Met-79→Thr and Arg-113→Lys) (22). In addition, both Bf1149 and Bf1150 strains harbored an identical 1,320-bp IS 8 bp upstream of the cfiA gene which differed from IS1186 and IS1168 by three and seven mismatches, respectively, including one (Glu-200→Lys) and two (Ala-178→Glu and Arg-46→Cys) missense mutations, respectively. IS1186 is one of the five elements reported at present to promote the expression of cfiA (19). IS1168 is 99.7% identical to IS1186 and has been identified upstream of 5-nitroimidazole resistance genes (11). Multiple isoforms of these ISs and various sites of insertion upstream of cfiA have been described previously (3, 19, 20, 24). Since Bf1149 and Bf1150 possessed strictly identical and original sequences encompassing the cfiA gene and the upstream IS1186/IS1168-related element, these isolates were concluded to be a single strain.

In conclusion, the cluster of four cases of B. fragilis bacteremia analyzed in this study appeared to include two independent endogenous cases (patients A and D) and two cross-transmitted cases (patients B and C). This is the first demonstration of B. fragilis cross-infection by genotypic typing methods. The sticking bandage fixing the IVP and the Huber needle is speculated to be the source of cross-transmission. This observation emphasizes the need for careful assessment of all items that come in contact with multiple patients.

This work was supported by grants from the Ministère de l'Education Nationale et de la Recherche (EA525) and the Programme de Recherche Fondamentale en Microbiologie et Maladies Infectieuses et Parasitaires (Réseau β-lactamase).