The study was performed in Moyen-Ogooué and Ngounié Provinces (combined population 162,000) in central Gabon; these 2 provinces cover an area of 56,285 km² and consist of predominantly dense primary rain forest. For the seroprevalence surveillance study, 304 serum samples were collected from healthy nonfebrile school children (12–20 years of age) living in and around Lambaréné, the capital of Moyen-Ogooué Province; these children also participated in a chemoprophylaxis study for malaria (15).

A prospective analysis of community-acquired bloodstream infections was performed at Albert Schweitzer Hospital (which admits ≈6,000 patients annually) in Lambaréné (population 24,000), in the Central African rain forest on the river Ogooué, Gabon. The rainy season starts in October and ends in June (including a short dry season in December–January). Mean annual rainfall is 1,981 mm (78 inches), which is equivalent to that in northeastern Thailand (16). Studies were approved by the Centre National de la Recherche Scientifique et Technologique, Libreville, and the scientific review committee of the Centre de Recherches Médicales de Lambaréné, Albert Schweitzer Hospital.

To obtain data about the prevalence and causes of community-acquired bloodstream infections in Lambaréné, we prospectively monitored all blood cultures for febrile patients admitted to Albert Schweitzer Hospital for 1 year (June 1, 2012–May 31, 2013) by using BacT/Alert PF (bioMérieux, Marcy l'Etoile, France). Criteria for ordering blood cultures were left to the discretion of the treating physician. Technicians and staff of the clinical microbiology laboratory received additional training on sample handling and processing (17,18). All oxidase-positive, gram-negative bacteria that were not Pseudomonas aeruginosa were further tested to determine whether they were B. pseudomallei by using the subculture and identification methods described below. Antimicrobial drug susceptibilities were determined by using Etest (bioMérieux) on Mueller-Hinton-agar (bioMérieux); when available, break points were defined as described (19).

During May 2012, presence and titer of antibodies to B. pseudomallei in healthy schoolchildren were determined using by the indirect hemagglutination assay (IHA) as described (20,21), with pooled antigens prepared from 2 B. pseudomallei isolates from Thailand. An antibody titer of ≥1:40 was used as the cutoff value for seropositivity (22).

During July 2012–September 2012, soil sampling to test for the presence of B. pseudomallei was based on consensus guidelines, and direct culture of soil in enrichment broth was performed (17,23). A total of 8 sites around the residences of children were selected on the basis of local maps and consultations with inhabitants throughout the provinces of Moyen-Ogooué (6 sites) and Ngounié (2 sites) and on known factors associated with the presence of B. pseudomallei (e.g., wet soil such as rice paddies or land use such as goat farming) (17) (Figure 1). Within each sampling area (50 × 50 m²), a fixed-interval sampling grid was used to collect 100 samples per field, 5 m apart. For each sample, 10 g of soil was collected from a depth of 30 cm, stored away from direct sunlight, and processed within 3 h.

Isolation of potential Burkholderia spp. from soil was performed as described (17,23). In brief, 10 g of soil was diluted in 10 mL of threonine–basal salt solution plus colistin at 50 mg/liter (TBSS-C50 broth) containing crystal violet and was vortexed for 30 s before incubation at ≈42°C for 48 h. Ten μL of supernatant was subcultured onto Ashdown-agar and incubated and examined every 24 h for 7 days. B. pseudomallei was identified by colony morphology, positive oxidase test result, inability to assimilate arabinose, antimicrobial drug susceptibility pattern (B. pseudomallei is generally resistant to gentamicin and colistin but susceptible to amoxicillin/clavulanic acid [1,2]), and results of API 20NE (bioMérieux) and B. pseudomallei–specific (Bps) latex-agglutination tests (18,24,25). Positive results were confirmed with molecular analysis. Soil type was determined by standard lithologic and pedologic analysis of sediments; for this purpose, 2 extra samples were collected per site from a depth of 30 cm (26). Sediment properties were compared with properties of other (typical) samples from the same locations as described in the recently published Soil Atlas of Africa (26).

Genomic DNA was extracted by using a DNeasy Blood and Tissue Kit (QIAGEN, Valencia, CA, USA) to perform multilocus sequence typing (MLST) (27). Primers used to amplify fragments of the 7 housekeeping genes were identical to those described at the Burkholderia MLST website (http://bpseudomallei.mlst.net/misc/info2.asp). For isolate B. thailandensis D50, the primer narK-up was replaced by narK-upAMC 5′-TCTCTACTCGTGCGCTGGGG-3′. Sequences of the 7 gene fragments of isolates from Africa were concatenated and combined with those from a selection of 971 sequence types (STs) representing all B. pseudomallei, B. mallei, and B. thailandensis isolates in the B. pseudomallei MLST database. Concatenated sequences were aligned and analyzed by using MEGA-6 (http://www.megasoftware.net). A phylogenetic tree was constructed by using a neighbor-joining algorithm and the Kimura 2-parameter model. Bootstrap testing was performed for 500 repetitions. Whole-genome sequencing was performed by using the MiSeq platform (Illumina, San Diego, CA, USA) as described (9).

Of the 941 bacterial blood cultures, 77 (8.2%) were positive for bacteria. The most prevalent isolate was Escherichia coli, responsible for 8 (10.0%) bloodstream infections, followed by Staphylococcus aureus (6 [7.8%]) and Salmonella enterica (6 [7.8%], 5 of which were nontyphoidal salmonellae). Other organisms that were isolated at least 5 times included Streptococcus pneumoniae (5 [6.5%]), Klebsiella pneumoniae (5 [6.5%]), and Enterobacter spp. (5 [6.5%]). B. pseudomallei was isolated from 1 (1.4%) patient, described in the case report.

A 62-year-old Gabonese woman was hospitalized in January 2013 with a 7-day history of fever, cough, weakness, headache, vomiting, and a painful knee. She did not report coughing or shortness of breath. She had poorly controlled diabetes mellitus and was taking glibenclamide. She had no history of cardiopulmonary or renal disease, was receiving no long-term medications other than glibenclamide, and did not smoke. She was a retired school teacher but still engaged in family farming. Physical examination revealed blood pressure of 160/90 mm Hg, a pulse rate of 130 beats per minute, and a temperature of 40.5°C. She had a wound with an underlying abscess on her right leg, together with diffuse tenderness of the right knee with warmth, erythema, and limitation of active and passive ranges of motion because of pain and effusion. Neurologic, cardiovascular, and respiratory examinations revealed no abnormalities. Laboratory findings obtained at admission showed an elevated blood glucose level of 24 mmol/L but values within reference range for creatinine (0.85 mg/dL), leukocytes (9,800 × 10³ mm³), and hemoglobin (9.2 g/dL). No other blood or urine test was performed, and chest radiographs were not taken. On hospitalization day 1, treatment with amoxicillin/clavulanic acid was empirically initiated for sepsis. On day 2, the abscess was incised and drained, and on day 3 antimicrobial drug therapy was switched to ceftriaxone. Cultures of blood, wound, and synovial fluid grew identical gram-negative rods, which were initially classified as Pseudomonas spp. No other pathogens were detected. The patient’s clinical condition deteriorated, and she died of septic shock on day 8. A postmortem examination was not performed.

After the patient’s death, the Pseudomonas species was classified as B. pseudomallei (patient strain Gb100) and confirmed by MLST and whole-genome sequencing. This isolate was later determined to be susceptible to trimethoprim/sulfamethoxazole, amoxicillin/clavulanic acid, ceftazidime, and meropenem (Table 1).

Of the 304 healthy schoolchildren for whom serum samples were tested for B. pseudomallei antibodies, 143 (47.0%) were male. Details for this cohort have been reported previously (15). For 43 (14.1%) children, an IHA titer was detectable; titers ranged from 1:10 to 1:80 (median 1:10, interquartile range 1:10–1:20). For 5 (1.6%) children, IHA titer was >1:40, which has been used as the cutoff value for seropositivity (22). None of the children had an IHA titer >1:160, which is considered by several centers in Thailand to support a diagnosis of melioidosis in patients with clinical features consistent with this diagnosis.

The predominant soil type in this area of Gabon was ferralsol, which is red and yellow weathered soil. The only exception was samples taken from a rice paddy near Mouila village, where the soil was gleysol (clay, a hydric soil saturated with groundwater long enough to develop a characteristic gleyic color pattern) (Table 2). B. pseudomallei was isolated from 21 (3%) of 800 soil samples taken from 3 (38%) of the 8 sample sites; the maximum number of positive samples for 1 site was 14 (14%) (Table 2). The biochemical profiles of all isolates were in accordance with B. pseudomallei (API 20NE code 1156576). The antibiogram of B. pseudomallei soil strain C2 is shown in Table 1.

The closely related B. thailandensis coexists with B. pseudomallei in the soil in Southeast Asia and Australia and is generally considered avirulent (5,28). We also identified B. thailandensis in the soil of Gabon (Figure 2). This strain, termed B. thailandensis soil strain D50, was positive by Bps latex agglutination. This B. thailandensis strain, API 20NE code 1157577, was susceptible to trimethoprim/sulfamethoxazole, amoxicillin/clavulanic acid, ceftazidime, and meropenem (Table 1).

The 3 isolates from Gabon contained previously described MLST alleles but belonged to novel STs. The patient isolate Gb100 (ST1127) and soil isolate C2 (ST1128) were single-locus variants and differed by 1 nt in the narK sequence only. Patient isolate Gb100 was also a single-locus variant of ST707 (single-nucleotide substitution in ndh). The only B. pseudomallei strain with ST707 in the database had been isolated in 2010 from a patient in the United Kingdom, 6 weeks after the patient had returned from a trip to Nigeria (12). The soil isolate C2 (ST1128) was a single-locus variant of ST7 (single-nucleotide substitution in ndh) and ST879 (single-nucleotide substitution in lipA). ST7 was represented by 2 isolates in the MSLT database, both isolated in 1963 from patients in Vietnam. The B. pseudomallei ST879 strain was isolated in 2011 from a patient in Spain, who had returned from a trip through Madagascar and 14 countries in West Africa (11). The soil isolate D50 (ST1126) was a single-locus variant of ST73. This ST is represented in the database by 2 B. thailandensis strains, 1 isolated from a foal in France and 1 isolated from the environment in Kenya. Phylogenetic analysis of the Gabon isolates together with 971 STs obtained from the MLST database by using the aligned concatenated sequences of the 7 loci in the neighbor-joining algorithm with the Kimura 2-parameter model showed that the patient isolate Gb100 and soil isolate C2, found near the community of the patient, grouped together with 7 STs. These 7 STs represented 10 B. pseudomallei strains isolated in Cambodia (2 strains), Vietnam (2 strains), Niger, Nigeria, Spain (imported), France (2 strains [1 imported]), and the United Kingdom (imported) (Figure 2). Again, patient isolate Gb100 and soil isolate C2 are most closely related to ST879. Soil isolate D50 grouped together with 3 STs representing 4 B. thailandensis strains isolated from Kenya, France, the United States, and Cambodia. Using this approach, we showed that the closest relatives of the strain that infected and eventually killed the patient reported here were ST879 and the strain isolated from soil around her community. Our whole-genome sequencing sample data have been submitted to a project that is undertaking whole-genome sequencing on a large number of B. pseudomallei isolates from around the world. This approach is anticipated to offer superior resolution of the global phylogeny of B. pseudomallei (9).