Burkholderia pseudomallei is an environmental Gram-negative bacillus and the cause of melioidosis. The clinical manifestations of melioidosis are broad and include disseminated disease with organ abscesses, severe sepsis, and mild infection of the skin and soft tissue [1]. Most patients have risk factors for infection, which include diabetes, heavy alcohol use, and chronic pulmonary or kidney disease [1]–[3]. The highest number of reported cases occurs in endemic regions of Thailand and Australia. Rising incidence rates have been recorded in northeast Thailand between 1997–2006, during which the average mortality rate was 42.6% [3]. In 2006, melioidosis and tuberculosis mortality rates in northeast Thailand were equivalent and second only to HIV/AIDS for infectious disease deaths [3]. In northern Australia the mortality rate over the last five years of the Darwin prospective melioidosis study was calculated at 9% [2]. The authors attributed the low mortality rate to early diagnosis and treatment, and access to and improvements in intensive care management [2].

Isolation of B. pseudomallei from clinical samples remains the “gold standard” against which other melioidosis diagnostics are compared [4]. Culture is routinely performed on multiple sample types (blood, urine, pus, sputum, etc.) and isolation of B. pseudomallei from any one of these cultures is diagnostic for melioidosis [5], [6]. However, recent modeling data has confirmed that culturing is an imperfect gold standard [7]. Furthermore, laboratory processing of positive samples takes 3–7 days [8]. This problem is compounded by the fact that many diagnostic laboratories may misidentify B. pseudomallei through lack of experience or validated diagnostic reagents [9]. Any delay in diagnostic confirmation is potentially important as B. pseudomallei requires therapy with ceftazidime or a carbapenem drug, which are not agents of choice for empirical therapeutic regimens. Taken together, these factors point to a clear need for a simple and rapid diagnostic test for accurate identification of B. pseudomallei directly on clinical samples or cultures.

Prior to diagnostic test development we identified a number of potential B. pseudomallei diagnostic biomarkers by In vivo Microbial Antigen Discovery (InMAD) [10], [11] that are shed or secreted and may be targeted to diagnose acute disease. Capsular polysaccharide (CPS) proved to be the most encouraging target; this molecule is a polymer of 1,3-linked 2-O-acetyl-6-deoxy-β-d-manno-heptopyranose residues [12]. We confirmed CPS was present in melioidosis patient serum and urine samples by antigen-capture ELISA utilizing a CPS-specific monoclonal antibody (mAb 3C5) [10]. The current report describes the characterization of mAb 3C5, quantification of CPS within patient samples, and optimization of the Active Melioidosis Detect lateral flow immunoassay (AMD LFI) for the rapid diagnosis of melioidosis.

Our previous report described the ability of mAb 3C5 to detect B. pseudomallei CPS in urine from patients with melioidosis [10]. Although encouraging, further experiments were required before constructing a point-of-care diagnostic assay to determine (i) the affinity of mAb 3C5 for CPS, (ii) the limit of detection of mAb 3C5 for CPS by ELISA, and (iii) the concentration range of CPS that accumulates in melioidosis patient samples.

SPR was used to determine the functional affinity of mAb 3C5 for B. pseudomallei CPS. Functional affinity is often referred to when describing the collective effects of mAb bivalency and antigen multivalency on binding (since CPS is composed of repeating epitopes). The functional affinity was evaluated on a BIAcore X100 sensor surface coated with immobilized CPS. The binding activity of mAb 3C5 was examined over a 60 s injection pulse. Total (resonance units) RU values were recorded following binding of a series of mAb concentrations (Fig. 1, left panel). These RU values were analyzed using a steady-state binding model (Fig. 1, right panel). This led to the calculation of a 50 nM dissociation constant (K_D) of mAb 3C5 for CPS. This is a relatively high affinity for a mAb specific to a polysaccharide antigen. This led us to expect that mAb 3C5 would perform well in an antibody-based detection assay.

An antigen-capture ELISA for CPS [10] was constructed to determine the limit of detection (LOD) that could be achieved with mAb 3C5 (Fig. 2). Due to the polyvalent nature of CPS, mAb 3C5 was used for both capture and detection in this assay. A two-fold serial dilution of mAb 3C5 was incubated in the solid phase of the 96-well microtiter plate vertically across all eight rows. Following a wash and blocking step, a two-fold serial dilution of purified CPS was incubated in the wells (horizontally). Captured CPS was detected with mAb 3C5 labeled with HRP. An optimal LOD of 0.2 ng/ml (2-fold over background) was achieved with a mAb 3C5 coating concentration of 2 µg/ml.

The antigen-capture ELISA was then used to quantify the amount of CPS within serum and urine samples collected from patients with culture-confirmed melioidosis in Thailand. Quantitative cultures were performed on urine samples prior to testing and are reported as CFU/ml (Table 2). Blood cultures were also tested although the CFU/ml was not determined. Each serum (isolated from blood) and urine sample was passed through a 0.22 µm filter in order to remove intact bacterial cells prior to shipment. In our previous report [10] we determined the highest fold dilution of these samples that yielded an ELISA OD₄₅₀ value ≥0.5. For the current study we were able to estimate the concentration of CPS within these samples by comparing the OD values to a standard curve produced with purified CPS (Table 2). CPS was detected in 6/10 filtered urine samples at concentrations ranging from 0.78–448 ng/ml. As expected, the concentration of CPS was higher in samples that contained more CFUs/ml. CPS was detected in all urine samples containing greater than 1.2×10⁴ CFU/ml. CPS was detected in 5/10 filtered serum samples at concentrations ranging from 0.85 to 6.7 ng/ml.

Following successful detection of CPS by ELISA, a prototype AMD LFI was constructed. A schematic of the components of the LFI is depicted in Fig. 3A. Initial LFI testing was performed on B. pseudomallei strain Bp82, a select agent excluded strain [16], and a strain of E. coli (negative control). Bp82 was not included in Table 1 since the strain was derived from B. pseudomallei strain 1026b, which is listed in Table 1. For each test, one single colony was collected with a sterile loop and resuspended in two drops of lysis buffer. The lysate was pipetted onto the LFI sample pad followed by addition of three drops of chase buffer. The fluid migrates by capillary action into the conjugate pad where gold-labeled mAb 3C5 binds to CPS present in the lysate. The gold-labeled mAb 3C5/CPS complex then migrates into the nitrocellulose membrane and is captured at the test line, which is unlabeled mAb 3C5 bound to the membrane. The absorbent or wicking pad allows for efficient capillary flow of the sample across the test line. The LFI used to analyze Bp82 showed test line and control line reactivity (Fig. 3B, top LFI) while the E. coli LFI was reactive only on the control line (Fig. 3B, bottom LFI). The tests are run for 15 minutes and results are recorded and imaged. Presence of the LFI control line ensures the test has run properly.

The LFI was tested for reactivity to B. pseudomallei and B. mallei in addition to other near neighbor species (Table 2). Strain panels tested included isolates selected by the Stakeholder Panel on Agent Detection Assay (SPADA) Burkholderia Working Group. The SPADA Burkholderia panel was compiled by a number of key stakeholders from federal agencies and biothreat researchers [17]. B. mallei has recently been shown to produce the identical manno-heptose capsule as B. pseudomallei [18]; we have previously shown mAb 3C5 reactivity to B. mallei CPS by Western blot [10]. The LFI testing was performed at the Centers for Disease Control and Prevention to evaluate analytical reactivity and specificity on inclusivity and exclusivity strain panels. Three colonies from each isolate listed in Table 1 were tested separately on the LFI. Of the B. pseudomallei isolates tested, 76/77 (98.7%) were positive; 30/33 (90.9%) of the B. mallei isolates were also positive. In addition, 35/36 (97.2%) of near neighbor species were negative by LFI. Eight Burkholderia thailandensis isolates were tested, and seven were negative. Other near neighbor species where also tested, including Burkholderia humptydooensis sp. nov., Burkholderia oklahomensis and Burkholderia cepacia complex (Bcc) species, all of which were negative. In addition, other medically relevant species of bacteria were negative for reactivity by Western blot (see footnote to Table 1) to mAb 3C5 (data not shown).

The LOD of the AMD LFI was determined to verify that the analytical sensitivity of the assay was sufficiently low to be used to detect CPS in patient samples. Purified CPS was tested on the LFI to determine the LOD under optimal conditions (Fig. 4). Dilutions of CPS were prepared in chase buffer and applied to the LFI sample pad. The LOD was estimated at or slightly below 0.2 ng/ml. In addition, purified CPS was spiked into control serum (Fig. 4B) and urine (Fig. 4C). Under these conditions the LOD was increased slightly when compared to dilution in chase buffer alone, however a clear reaction was apparent at 0.2 ng/ml.

The ability of the AMD LFI to accept a variety of patient samples was assessed with a limited number of culture-positive melioidosis samples in Australia. These samples were also used to optimize sample preparation for the AMD LFI. The LFI was designed to accept multiple sample matrices, which is critically important for the diagnosis of melioidosis. As shown in Fig. 5A the samples tested included serum, urine, sputum, pus and pleural fluid collected from culture-confirmed melioidosis patients (samples were not collected from the same patient) in Thailand and Australia. Preparation of each sample prior to application to the sample pad is described in the Methods section. The melioidosis patient urine samples that were tested by antigen-capture ELISA (Table 2) were also tested by LFI (Fig. 5B). The urine samples that were positive by ELISA were also positive by LFI. Qualitatively, the test line intensity of the positive urine samples was congruent with their corresponding ELISA values.

A number of assays have been developed to diagnose melioidosis prior to culture results becoming available. PCR has been developed but is not in routine practice because it is limited by low sensitivity, most likely stemming from the low concentration of B. pseudomallei in blood and the co-purification of PCR inhibitors with target DNA [19]–[21]. However, a recently developed Type III secretion system (TTS-1) real-time PCR assay has been shown to be superior to previously developed PCR assays for detection of B. pseudomallei DNA in clinical specimens [21]. When compared to culture the TTS-1 assay had a sensitivity and specificity of 80% and 100%, respectively. The indirect hemagglutination assay (IHA) is a rapid and inexpensive method used to detect antibodies produced during infection that are specific to B. pseudomallei. However, a large percentage of healthy individuals in endemic areas are seropositive [22], [23]. This point is underscored by the fact that nearly 70% of children in northeast Thailand are seropositive for B. pseudomallei antigens [24], [25]. Consequently, the IHA (or any serological test for melioidosis) has limited clinical utility in the endemic setting [1], [26].

Antigen detection by immunofluorescence assay (IFA) or latex agglutination is commonly used in endemic areas. IFA is used in northeast Thailand for rapid diagnosis directly from patient samples containing high levels of B. pseudomallei (sputum, pus, urine and respiratory secretions) [27], [28] and from blood cultures [29]. The main drawback of IFA is the requirement for a fluorescent microscope and the requisite expertise, which is not feasible in most endemic settings. In addition, although specificity of the IFA is high, the sensitivity has recently been determined to range from 45–48% when used directly on clinical samples [27]. Latex agglutination is an inexpensive technique that is effective at identifying B. pseudomallei from cultures of patient samples grown on agar plates or within liquid broth [30]–[33]. The agglutination assay is able to detect B. pseudomallei at concentrations of 1–2×10⁶ CFU/ml; this limits its utility to cultured patient samples or colonies isolated on solid agar [32], [33].

Our LFI is similar in design to those currently used for the diagnosis of Streptococcus pneumoniae and Legionella pneumophila [34]. The L. pneumophila assay is a first-line test that relies on detection of antigen produced by the bacterium within patient urine [35]. We anticipate the AMD LFI can also be used as a first-line test and offer an improvement over the current rapid techniques for the diagnosis of melioidosis. In addition we believe lateral flow devices are well suited for resource poor settings in that they are inexpensive, rapid, sensitive, and stable at room temperature. In addition, LFIs do not require expensive equipment and they can accept multiple sample matrices, two characteristics that are essential for the diagnosis of melioidosis in resource poor settings.

IgG3 mAb 3C5 possesses many important characteristics that are necessary for the development of an antigen detection assay. It has a relatively high affinity for its target antigen and shows acceptable analytical reactivity and specificity. The high affinity translates into a lower limit of detection for CPS by ELISA and LFI. Interestingly, the LFI had a comparable analytical sensitivity to the ELISA (∼0.2 ng/ml) when CPS was diluted in chase buffer. The analytical sensitivity was slightly lower when CPS was spiked into control serum and urine. When tested by LFI, 98.7% of B. pseudomallei isolates were positive while 97.2% of near neighbor species were negative. Both the false-negative and false-positive LFI results can be explained through sequencing analysis. The one isolate that produced a false negative (MSHR1655) originated from a patient that developed a persistent asymptomatic B. pseudomallei infection in Australia. A frameshift mutation was identified within the wcbR gene of this isolate [36]. A B. pseudomallei strain (K96243) with a wcbR mutation was recently shown to have greatly reduced CPS expression [37]. The one B. thailandensis isolate that produced the false positive had been previously shown to encode the CPS biosynthetic operon [38], [39].

An essential aspect of the current study was the quantification of CPS within patient samples. This was accomplished by comparing ELISA values generated from patient samples with a standard curve generated with known concentrations of purified CPS. Over half of the filtered serum and urine samples from melioidosis patients had levels of CPS within the detection range of the AMD LFI. The LFI detected CPS in 6/10 culture-positive urine samples from melioidosis patients. We anticipate that if the urine had not been filtered more of the samples would have been positive. Patient serum samples were not tested on the LFI due to insufficient volumes, but half contained concentrations of CPS (as determined by ELISA) that could be detected by the AMD LFI. This is encouraging since the mean concentration of B. pseudomallei in patient blood is ∼1 CFU/ml [5], [6]. We anticipate that CPS may be shed from internal abscesses into the blood; so theoretically, even if the concentration of bacteria in blood is low, the concentration of CPS may be within the detectable range of the LFI. CPS could not be detected in filtered urine samples that contain low levels of bacteria, suggesting that CPS may not be shed into urine to detectable levels from the blood.

This study describes the development and optimization of a prototype LFI for the rapid diagnosis of melioidosis, including protocols for the preparation of different sample types. This is essential since the LFI will be used to test at least four different bodily fluids, bacterial colonies grown on solid agar, and bacterial liquid cultures from patient samples. We anticipate routine testing can be performed on all patient sample types, and the clinical sensitivity of the LFI will be related to the specific sample type tested. The sample type producing the lowest sensitivity will most likely be blood; this is related to the low levels of B. pseudomallei found in this sample type [5], [6]. However, we believe when the LFI is used to test urine, sputum, and pus, high sensitivity will be achieved due to the increased CFU/ml values in these matrices. Now that we have developed reliable sample preparation guidelines we will perform a larger preclinical analysis in the endemic areas of Thailand and Australia. The preclinical analysis will compare the performance of the LFI with the TTS-1 real-time PCR assay, IFA, and culture (the current “gold standard” for diagnosis of melioidosis). This will allow us to determine clinical sensitivity and specificity and the diagnostic utility of the assay. Further studies are underway to isolate additional CPS specific mAbs that possess higher affinities than 3C5. Incorporation of such mAbs into the AMD LFI may lead to increased analytical and clinical sensitivity.