Burkholderia pseudomallei Misidentified by Automated System

After returning from Thailand, a 35-year-old man from Switzerland was hospitalized with an abscess of the head. Material cultured from the abscess and adjacent bone grew a gram-negative rod, which was misidentified by an automated microbiology system as Burkholderia cepacia. The organism was eventually identified by molecular methods as B. pseudomallei.

B urkholderia pseudomallei, the etiologic agent of melioidosis, can cause pyogenic or granulomatous lesions and is endemic to tropic regions, mainly in Southeast Asia and northern Australia. This organism is a potential category B bioterrorism agent (1). Melioidosis occurs sporadically in travelers returning from disease-endemic areas, and laboratories in regions where this disease is not endemic are not familiar with identifi cation of B. pseudomallei, thus potentially leading to misidentifi cation (2). We report the misidentifi cation of this organism by an automated microbiology system.

The Study
On August 20, 2008, a 35-year-old man from Switzerland was admitted to the Cantonal Hospital in St. Gallen, Switzerland. He had extradural cranial abscess of the right parietal area and a defect in adjacent bone. In May and June 2008, he traveled to Singapore and Malaysia (Kuala Lumpur and the Perhentian Islands), then to southwestern (Ko Samui, Ko Tau) and northern (Chiang Mai) Thailand where he went trekking and river rafting.
The patient did not remember receiving a head injury during his trip. Seventeen days after returning to Switzerland, he had a swelling in the right parietal area of the head. The parietal bulge increased, but puncture by his general practitioner showed no aspirate. During the next 7 weeks, the bulge became painful and secretion of pus was noted at the time of admission.
At admission, his general condition was good and he had no signs of systemic infl ammation. He did not have any neurologic defi cits or other abnormal fi ndings. Test results for complete blood cell count, C-reactive protein and creatinine levels, and liver functions were normal. Computed tomography and magnetic resonance imaging of the head showed the abscess and a small defect of bone (Figure). The abscess and part of the cranial bone were then surgically removed.
Culture material from the abscess and biopsy specimens of the abscess capsule and the cranial bone grew gram-negative, oxidase-positive rods, and smooth creamy colonies on sheep blood agar after incubation for 48 hours at 35°C. For identifi cation, a suspension of the isolate was prepared and tested in a UNMIC/ID-62 panel of the BD Phoenix Automated Microbiology System (Becton Dickinson AG, Allschwil, Switzerland) per the manufacturer's instructions. This system identifi ed the isolate as B. cepacia (99% confi dence). When a species is identifi ed with >90% confi dence, the Phoenix System gives an identifi cation result as a measure of likelihood that the identifi cation is the only correct one.
The patient was discharged from hospital after 5 days with a preliminary diagnosis of B. cepacia infection of soft tissue (the cranial bone lesion was attributed to trauma). He was treated with oral cotrimoxazole (160/800 mg 2× a day) for 16 days. The isolate was sensitive to cotrimoxazole (MIC = 1 mg/L for trimethoprim and 19 mg/L for sulfamethoxazole) by the Phoenix System for B. cepacia (Becton Dickinson). The isolate was sensitive to imipenem, ceftazidime, doxycycline, cotrimoxazole, and tetracycline by Etest (AB Biodisk, Solna, Sweden). The diagnosis was regarded as preliminary for the following reasons. First, identifi cation of B. cepacia by common automated identifi cation instruments such as the Phoenix System or VITEK 2 (bioMérieux, Geneva, Switzerland) requires confi rmatory identifi cation by molecular tests (3). Second, an abscess is an uncommon location for B. cepacia (4). Third, the bacterial colonies emitted an unexpected, earthy odor. Fourth, the isolate was sensitive to amoxicillin (MIC = 8 mg/L) and clavulanate (MIC = 4 mg/L).
To verify identifi cation of the isolate, a 500-bp fragment of the 16S rRNA gene was amplifi ed and sequenced by using the Fast MicroSeq 500 16S rDNA Bacterial Identifi cation Kit and a PRISM 310 Genetic Analyzer (both from Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. Sequence analysis was performed by using MicroSeq ID Microbial Identifi cation Software (Applied Biosystems). The MicroSeq ID 2.0 500bp library identifi ed B. pseudomallei (ATCC 23343, gb DQ108392.1) with a 446-bp consensus length and 1 mismatch with B. mallei (NCTC 10247, gb CP000548.1). These results suggested that our isolate was B. pseudomallei.
Additional investigations with a fl agellin C gene-specifi c real-time PCR (5) suggested that the isolate was B. pseudomallei or B. mallei. Multilocus sequence typing (6) showed the allelic profi le 1/2/3/1/1/2/1, which identifi ed the isolate as B. pseudomallei sequence type 306. This sequence type was isolated from serum samples of 2 patients in Thailand with invasive melioidosis in 1991 and 2006.
Forty-four days after surgery, new biopsy specimens of soft tissue below the parietal scar and a sample of galea aponeurotica were cultured for B. pseudomallei after the patient received 2 weeks of inadequate therapy with lowdose cotrimoxazole to determine its drug resistance pattern. Cultures yielded only Staphylococcus epidermidis, which was regarded as a contaminant. To ensure eradication of B. pseudomallei, the patient was treated with imipenem (500 mg, 4×/day), cotrimoxazole (400 mg trimethoprim/ day), and leucovorine (15 mg, 3× a week) for 2 weeks and later with cotrimoxazole (320 mg trimethoprim/day) and leucovorine (15 mg, 3×/week) for 6 months. The patient recovered after 6 months; he had a small indentation without signs of infl ammation at the site of the abscess.
For comparison with the Phoenix System, we retrospectively analyzed the isolate by using the API 20NE biochemical test panel V7.0 (bioMérieux). This panel identifi ed the isolate as B. pseudomallei (profi le 1156577; 99.9% ID, 1.0 T).

Conclusions
Automated methods for identifi cation of bacterial isolates and testing of antimicrobial drug susceptibility, such as the Phoenix System, have become standard in most clinical laboratories because they are easy to use and turnaround time is rapid. The Phoenix System uses fl uorogenic and chromogenic substrates for its identifi cation algorithms, a broth-based antimicrobial drug-susceptibility testing method, and a data-processing application (Phoenix EpiCenter; Becton Dickinson AG). Unfortunately, B. pseudomallei was not in the database of this system, which led to misidentifi cation of our isolate as B. cepacia. Failure to correctly identify B. pseudomallei has also been reported with another widely used automated system, the Vitek2 system (7). To identify isolates from patients traveling to disease-endemic areas, automated systems should be updated for identifi cation of B. pseudomallei and other rare bacteria that cause severe human infections, are hazardous to laboratory personnel, and may be used as bioterrorism agents (e.g., Brucella spp. and B. mallei).
Three modes of acquisition (inhalation, ingestion, and inoculation) are recognized for B. pseudomallei infection. Skin and soft tissue infections may occur after minor wounds or from hematogenous spread (8). In our patient, inoculation of the skin with B. pseudomallei after a minor injury in Thailand is probably the mode of infection.
Although laboratory personnel handled cultures of B. pseudomallei for identifi cation and drug resistance testing and smelled culture plates without knowing the isolate's identity, none became ill or showed signs of melioidosis. According to expert consensus (9), exposure of our laboratory personnel was classifi ed as a low risk. Serum samples were stored at -30°C to enable serologic testing for any subsequent illness.
Laboratories in regions where B. pseudomallei is not endemic should be aware of misdiagnosis of isolates by automated methods for bacterial identifi cation and antimicrobial drug susceptibility testing. Identifi cation of Burkholderia spp. by the Phoenix EpiCenter should be confi rmed by molecular methods and by the API 20NE system in suspected cases of B. pseudomallei infection. Because of its high rate of accuracy and ease of use, the API 20NE system should be used fi rst for any suspected colony when automated systems do not contain the adequate profi le (10).