Gas chromatography–mass spectrometry method for rapid identification and differentiation of Burkholderia pseudomallei and Burkholderia mallei from each other, Burkholderia thailandensis and several members of the Burkholderia cepacia complex
Department of Chemistry and Biochemistry, Brigham Young University, Provo, UT, USA
To develop a simple gas chromatography–mass spectrometry (GC-MS) method for the detection and differentiation of Burkholderia pseudomallei and Burkholderia mallei from each other, Burkholderia thailandensis and several members of the Burkholderia cepacia complex.
Methods and Results
Biomarkers were generated by one-step thermochemolysis (TCM) and analysed using a GC-MS system. Fragments of poly-3-hydroxybutyrate-co-hydroxyvalerate [poly(3HBA-co-3HVA)] produced by TCM were useful biomarkers. Several cellular fatty acid methyl esters were important in differentiating the various Burkholderia species. A statistical discrimination algorithm was constructed using a combination of biomarkers. The identities of four B. pseudomallei strains, four B. mallei strains and one strain of each near neighbour were confirmed in a statistically designed test using the algorithm. The detection limit for this method was found to be approximately 4000 cells.
The method is fast, accurate and easy to use. The algorithm is robust against different growth conditions (medium and temperature).
Significance and Impact of the Study
This assay may prove beneficial in a clinical diagnostic setting, where the rapid identification of B. pseudomallei is essential to effective treatment. This method could also be easily employed after a biological attack to confirm the presence of either B. pseudomallei or B. mallei.
Several species within the genus Burkholderia are classified as human pathogens. Perhaps the most well known of these pathogenic species are Burkholderia pseudomallei, Burkholderia mallei and several members of the Burkholderia cepacia complex. Burkholderia pseudomallei is a soil saprophyte that is indigenous to South-East Asia, Northern Australia and tropical regions near the equator (Howe et al. 1971; Chaowagul et al. 1989). It is also the causative agent of melioidosis, a human disease with symptoms that range from localized abscesses to acute septicaemia and pneumonia. Melioidosis has a mortality rate of 40% in Northern Thailand, 19% in Northern Australia and 39% in Singapore (Currie et al. 2000; Cheng and Currie 2005; Limmathurotsakul et al. 2010). In fact, during the first 48 h of hospital admittance, untreated cases of septicaemia have mortality rates as high as 80–90% (White et al. 1989; Sanford 1995). Additionally, in north-east Thailand, melioidosis accounts for 20% of community-acquired bacteremia and is the third most common cause of death by an infectious disease, following AIDS and tuberculosis (Suputtamongkol et al. 1994; Limmathurotsakul et al. 2010). The number of cases of melioidosis is also increasing in other populated countries such as Taiwan, China, Brazil and India (John et al. 1996; Yang 2000; Currie et al. 2008; Chen et al. 2010; Brilhante et al. 2012).
Burkholderia mallei is the causative agent of glanders, an abscess-forming infection predominantly found in the equine population. Although rare, it can cause serious disease in humans. Veterinarians, laboratory workers, equine handlers and slaughterhouse workers are at risk due to the possibility of repeated exposure to the micro-organism. Humans infected with B. mallei experience fever, rigours, malaise, diaphoresis, pneumonia, bacteremia, pustules and abscesses (Gregory and Waag 2007; Whitlock et al. 2007). The disease has a 95% case fatality rate for untreated septicaemia and a 50% case fatality rate in antibiotic-treated patients (Sanford 1995; Whitlock et al. 2007; Spickler 2008). Without antibiotic treatment, death typically occurs in 7–10 days (Gregory and Waag 2007; Whitlock et al. 2007).
Burkholderia pseudomallei and B. mallei are very closely related to each another. This has been demonstrated in part by comparing their cellular lipid and fatty acid compositions (Rogul et al. 1970; Yabuuchi et al. 1992; Godoy et al. 2003). In addition, it has been shown that these organisms have a high degree of genetic similarity that has created difficulties in developing accurate molecular-based differentiation assays (Sprague et al. 2002; Gee et al. 2003; Lee et al. 2005). Both organisms are recognized by the Centers for Disease Control and Prevention as possible bioterrorism agents based on their low infectious dose and their potential to cause widespread disease.
Burkholderia thailandensis is also closely related to B. pseudomallei and was only recently classified as a new species based on differences in 16s rRNA sequences, some biochemical properties and a much lower virulence in humans (Brett et al. 1998). There are very few cases of human infection by B. thailandensis, and those presumably resulted from a high infectious dose (Dharakul et al. 1999; Lertpatanasuwan et al. 1999; Glass et al. 2006). Although B. thailandensis holds very little clinical significance, it does share several virulence homologues with B. pseudomallei and B. mallei and is thus considered by many to be a model organism for studying the pathogenesis of various Burkholderia species. In addition, B. thailandensis is known to colocalize with B. pseudomallei in the environment and has similar phenotypic characteristics to that of B. pseudomallei by routine diagnostic tests (Thibault et al. 2004).
Burkholderia cepacia, Burkholderia multivorans, Burkholderia vietnamiensis and several other closely related bacterial species make up the Burkholderia cepacia complex (Vandamme and Dawyndt 2011). These organisms were originally believed to be only plant pathogens, but later emerged as important opportunistic pathogens causing chronic and life-threatening respiratory tract infections in patients with cystic fibrosis (Lipuma 2005). These members of the B. cepacia complex share many genetic similarities with B. pseudomallei, B. mallei and B. thailandensis.
The similarities between these different Burkholderia species have complicated the design and development of detection and differentiation assays. Nonetheless, several methods have been developed for the identification and discrimination of various combinations of the Burkholderia species that were described above. These methods include serological tests (Samosornsuk et al. 1999; Steinmetz et al. 1999; Anuntagool et al. 2000; Chenthamarakshan et al. 2001; Cheng et al. 2006), biochemical tests (Inglis et al. 1998; Lowe et al. 2002; Glass and Popovic 2005; Amornchai et al. 2007), microscopic methods (Walsh et al. 1994; Wuthiekanun et al. 2005; Hagen et al. 2011), polymerase chain reaction (PCR) assays (Bauernfeind et al. 1999; Mahenthiralingam et al. 2000; Thibault et al. 2004; Novak et al. 2006; Bowers et al. 2010; Puthucheary et al. 2012) and the gas chromatographic analysis of cellular fatty acids (Inglis et al. 2003, 2005). However, there are limitations associated with each of these assay types.
In endemic areas, serological tests for B. pseudomallei are unreliable due to the frequent seroconversion of individuals previously exposed to the organism (White 2003). Therefore, these serological tests have low sensitivity and specificity in areas of endemicity, but may prove useful in nonendemic areas (Wuthiekanun et al. 2004; Cheng et al. 2006). Biochemical assays have frequently misidentified B. pseudomallei as Pseudomonas spp., B. vietnamiensis, Stenotrophomonas maltophilia and Chromobacterium violaceum (Inglis et al. 1998; Lowe et al. 2002; Glass and Popovic 2005). Commercial biochemical tests have also misidentified members of the B. cepacia complex as Burkholderia gladioli, Ralstonia pickettii, Alcaligenes spp., Sten. maltophilia, Flavobacterium spp. and Chryseobacterium spp. (Kiska et al. 1996; McMenamin et al. 2000). In addition, many of these detection methods require the organism to be cultured prior to testing, which may take up to 7 days. Furthermore, similar colony morphologies and biochemical functions make it difficult to differentiate between the different species of Burkholderia (Wongtrakoongate et al. 2007; Chantratita et al. 2008).
PCR has revolutionized microbial detection due its accuracy, sensitivity and speed. PCR-based assays are usually designed around a well-conserved gene. However, it is possible, especially in emerging pathogens, for mutations to occur in the gene of interest, which can compromise the assay (Klein 2002). Restricting bacterial detection to a single target is another limitation of PCR-based assays. This issue can be overcome by developing multiplex PCR assays; however, such assays are difficult to implement due to the high degree of optimization that is required. The high sensitivity of PCR-based assays is also a limitation. False positives can arise from background contamination from external sources of DNA, such as the ‘carry-over’ products from earlier PCRs (Fredricks and Relman 1999; Yang and Rothman 2004). Conversely, false negatives can occur due to inadequate removal of PCR inhibitors. Although PCR-based assays have the potential to provide high throughput, the limitations associated with possible gene mutations, false positives, false negatives, sample processing and the need to validate with other established assays decrease the overall throughput of the entire PCR process.
Species-specific cellular fatty acids are important chemical markers that are frequently used in bacterial taxonomy and classification (Vandamme et al. 1996). A gas chromatography (GC)-based commercial microbial identification system was developed by MIDI (Newark, DE, USA) and provides a database of bacterial cellular fatty acid profiles. However, this method has several serious limitations. First, the organism must be cultured under standardized growth conditions. This requirement stems from the fact that fatty acid composition is influenced by growth medium and incubation temperature (Krejci and Kroppenstedt 2006). Second, sample preparation for this method includes several time-consuming and complicated steps (i.e. saponification, methylation, extraction and washing). These limitations indicate that the method is insufficiently robust and not suited for field detection. However, using this commercial method, Krejci and Kroppenstedt (2006) described certain fatty acids that are unique to the Burkholderia species and that could potentially be used for their detection and differentiation. For example, they proposed that the C16:0 3-OH fatty acid could be used for the identification of the B. cepacia complex.
Poly(3-hydroxyalkanoates) (P3HAs) might also aid in the detection and differentiation of Burkholderia species. They represent a class of biodegradable thermoplastics that are synthesized by a wide variety of bacteria (Steinbuchel 1991; Steinbuchel and Valentine 1995). Poly-3-hydroxybutyrate (P3HB), poly-3-hydroxyvalerate (P3HV) and poly-3-hydroxybutyrate-co-3-hydroxyvalerate [poly(3HB-co-3HV)] are the most extensively studied polyhydroxyalkanoate (PHA) compounds. These compounds are typically used by bacteria as a reserve carbon and energy source during unfavourable growth conditions (Steinbuchel and Valentine 1995). In addition, poly(3HB-co-3HV) has superior mechanical properties that have drawn increasing interest from bio-industrial fields that require biodegradable and biocompatible materials (Kim do et al. 2009). Burkholderia species are capable of accumulating poly(3HB-co-3HV) from various carbon sources, including glucose, fructose, acetate, glycerol and lactase (Kim do et al. 2009). We hypothesized that P3HB might be a suitable identification target for discrimination between the different species of Burkholderia.
In this study, we developed a simple method for the rapid identification and differentiation of B. pseudomallei and B. mallei from each other, B. thailandensis and several members of the B. cepacia complex (B. cepacia, B. multivorans and B. vietnamiensis) using a gas chromatography–mass spectrometry (GC-MS) system. The method relied on several cellular fatty acids and poly(3HBA-co-3HVA) derivatives as biomarkers. These biomarkers were released from the bacterial cells and derivatized into compounds that were more amenable to GC analysis via a single-step thermochemolysis (TCM) procedure. The TCM procedure was a modified version of the protocol that was used to generate biomarkers from the spores of several Bacillus species (Li et al. 2012). A statistical discrimination algorithm was then constructed using a combination of biomarkers, and the identities of the different species of Burkholderia were confirmed in a statistically designed test using the algorithm. These results demonstrated that the algorithm was robust against different growth conditions (i.e. medium and temperature).
Materials and methods
HPLC grade methanol (MeOH) and sulfuric acid (H2SO4) were purchased from Sigma Chemical (St. Louis, MO, USA). Hydrogen methyl sulfate (HMeSO4) was chosen as the derivatization agent because, as shown in past studies, it performed best in terms of cost, speed, safety and GC response (Antolin et al. 2008). It was prepared as a 20% solution of H2SO4 in methanol (v/v). The solution was allowed to rest at room temperature for approximately 8 days prior to use, which gave the reagents enough time to react. Phosphate buffer (1 mol l−1, pH 6·5) was made by mixing 1 mol l−1 potassium phosphate monobasic solution and 1 mol l−1 potassium phosphate dibasic solution until the pH reached 6·5; both reagents were purchased from Sigma Chemical.
The bacterial strains used in this study are described in Table 1. Strains of B. pseudomallei and B. mallei were selected as representatives of the major genetic clades (U'Ren et al. 2007). This was done to represent the extant genetic diversity within the two species. The near neighbours were selected based on their genetic similarity to the virulent strains (Levy et al. 2008). The purity of each strain was verified by Gram stain and subsequent inspection using a light microscope. The identities of the B. mallei and B. pseudomallei strains were further confirmed by the GC analysis of cellular fatty acids using an Agilent 6890 Series Gas Chromatograph (Santa Clara, CA, USA) and software purchased from MIDI (Newark, DE, USA). This method of identification was also attempted in conjunction with the other species of Burkholderia. Furthermore, a real-time PCR assay was used to definitively confirm the identity of each B. mallei and B. pseudomallei strain (U'Ren et al. 2005). The assay was conducted as described by U'Ren et al. (2005) and using primers from Integrated DNA Technologies (Coralville, IA, USA) and TaqMan® probes from Applied Biosystems (Foster City, CA, USA).
Table 1. Burkholderia strains used in this study
PHLS, Public Health Laboratory Service, UK; NCTC, National Collection of Type Cultures, UK; ATCC, American Type Culture Collection, USA; CDC, Centers for Disease Control and Prevention, USA.
Procedures involving B. mallei and B. pseudomallei were performed under biosafety level 3 (BSL-3) operating conditions. All other procedures were conducted under biosafety level 2 (BSL-2) operating conditions.
Each organism was initially cultured on blood agar that contained a Columbia Blood Agar Base (Becton, Dickinson and Company, Sparks, MD, USA) and 5% (v/v) sheep blood (Hema Resource and Supply, Aurora, OR, USA). After incubating for approximately 2–3 days at 37°C, isolated colonies were spread over blood agar and brain heart infusion agar (Becton, Dickinson and Company). All organisms were cultured on both types of growth medium and grown at two different temperatures (32 and 37°C) for approximately 2–3 days. These conditions represent reasonable and accessible media types and appropriate temperatures for culturing members of the Burkholderia genus.
The confluent bacterial growth was transferred to 5 ml of physiological saline solution and pelleted via centrifugation at 3200 g for 5 min at room temperature. Subsequently, the pellet was suspended in 5 ml sterile HPLC water (Sigma Chemical). An aliquot (1 ml) of this cell suspension was then transferred to a 1·7-ml microcentrifuge tube (BioExpress, Kaysville, UT, USA) and pelleted by centrifugation at 20 000 g for 5 min at room temperature.
The cell pellet was suspended in methanol to give a final concentration of approximately 1 × 107 cells ml−1. A 20-μl aliquot of the cell suspension was then transferred to a 1·7-ml microcentrifuge tube and combined with 20 μl of HMeSO4. After thoroughly mixing the suspension with a pipette, aliquots of 20 μl were transferred to clear flat-bottomed crimp vials (7 × 40 mm; National Scientific, Rockwood, TN, USA). The vials were capped with crimp top seals (8 mm, clear PTFE/red rubber, National Scientific) and sealed using a crimping tool. A custom-machined heating block and a digital mini temperature CSC 32 controller (Omega, Stamford, CT, USA) were used to heat the vials to 140°C for 5 min. The heating block had enough slots to accommodate four glass vials at one time. After returning the vials to room temperature, they were stored at −20°C until the viability testing or GC/MS analysis was conducted. Two vials were prepared for every sample that originated in the BSL-2. Alternatively, four vials were prepared for every sample that originated in the BSL-3, two of which were subjected to a viability testing.
Vials were injected with 200 μl of phosphate buffer (1 M, pH 6·5) to neutralize the acidic solution. This was done using 1-ml syringes and 22-gauge 1 1/2-inch needles that were purchased from Becton, Dickinson and Company. The vials were then mixed vigorously with a vortex mixer for 5 s. The entire contents of the vials were cultured on blood agar plates. Subsequently, the vials were discarded, and the plates were incubated for 5 days at 37°C. If no growth was observed, then sterility was confirmed, and the remaining vials were removed from the BSL-3 facility for GC-MS analysis.
Vials were injected with 400 μl phosphate buffer (1 mol l−1, pH 6·5) to neutralize the acidic solution. This was done using 1-ml syringes and 26-gauge 1 1/2-inch needles that were purchased from Becton, Dickinson and Company. Biomarkers were extracted using a 2-cm divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS) solid-phase microextraction (SPME) fibre that was purchased from Supelco (Bellefonte, PA, USA). The SPME fibre was inserted through the rubber septum of the capped vial (Li et al. 2012). Roughly, half of the fibre was submerged in the liquid phase, and half was exposed to the vapour phase. This allowed sampling of biomarkers that favoured partitioning in both vapour and liquid phases (Li et al. 2012). The extraction was conducted at 80°C for 10 min. The SPME fibre was then placed in the GC injection port for 2 min for desorption.
An Agilent 6890 GC was used in the analysis of all samples. The GC was equipped with a split/splitless injector containing a 79-mm-long × 1·2 mm i.d. × 6·3 mm o.d. deactivated fused silica liner (SGE, Austin, TX, USA) and a 30 m × 0·25 mm i.d. × 0·25 μm Zebron-FFAP column from Phenomenex (Torrance, CA, USA). The injection port was set at 260°C for all samples, and the inlet was operated in the splitless mode at a constant helium pressure of 24 psi. The temperature programme was set from 60°C (1-min initial hold), increased at 10°C min−1 to 170°C (1-min hold), then 5°C min−1 to 210°C (1-min hold) and finally 8°C min−1 to 250°C (5-min hold). The GC was coupled to an Agilent 5973 MS with electron ionization source and quadrupole mass analyser set to scan over a 33–550 m/z range. The transfer line to the MS was maintained at 230°C. Blanks (SPME fibre without sample extraction) were performed at the beginning of each day and periodically during the analyses. Cleaning the SPME fibre in between runs was achieved by placing the fibre in deionized water with agitation for approximately 20–30 min. The SPME fibre was then inserted into the GC-MS injection port for 1 min to remove any possible contamination or carry-over from previous samples. ChemStation software package (Agilent Technologies, Santa Clara, CA, USA) was used for GC-MS data analysis.
The experimental design entailed three batches of experiments, which were separated by more than a month. The first batch was used for constructing and training the differentiation algorithm. In this batch, each of the four B. pseudomallei strains and the four B. mallei strains was cultured under each of the four growth conditions and replicated twice. In addition, each of the four near neighbours was cultured under each of the four growth conditions; however, they were replicated three times. In total, the first batch included 32 samples of B. pseudomallei, 32 samples of B. mallei and 48 samples of the near neighbours. Construction of the algorithm was made using ‘R’, a statistical software package. The second and third batches were used for validating and testing the algorithm, respectively. In these batches, all of the organisms were cultured under each of the four growth conditions and were replicated twice. These batches each contained 32 samples of B. pseudomallei, 32 samples of B. mallei and 32 samples of the near neighbours. The algorithm was applied to the test data set in a blind manner. The classification of each tested vial was made by the computer algorithm using these data without knowledge of what species or how many samples of each species were in the data set. Operation of the algorithm was performed without input from the experimenters. The test data set was never used in algorithm construction.
Biomarker identification and confirmation
Fatty acid methyl esters (FAMEs) were named and their structures predicted by searching the Mass Spectral Library that is available through the National Institute of Standards and Technology (NIST). The majority of the FAMEs were previously described in the literature, which allowed us to further confirm their names and structures (Phung et al. 1995; Krejci and Kroppenstedt 2006; Levy et al. 2008). In addition, the identities of the biomarkers generated from poly(3HBA-co-3HVA) were confirmed by subjecting a poly(3HBA-co-3HVA) standard (Sarchem Laboratories, Farmingdale, NJ, USA) to the TCM procedure and subsequent GC-MS analysis.
A suspension of B. thailandensis cells in methanol was prepared as described above. The concentration of the bacterial cells in the suspension was determined using a Petroff–Hausser counting chamber (Hausser Scientific, Horsham, PA, USA). The suspension was then serially diluted in methanol to extinction. Subsequently, two samples from each dilution were prepared, processed and analysed as described above. This procedure was repeated on four different occasions. Burkholderia thailandensis was used to estimate the detection limit because it displayed very low abundance of a key biomarker (2-BAME).
Biomarker identification and confirmation
The TCM procedure generated biomarkers from the cellular fatty acids and the poly(3HBA-co-3HVA) polymers that were found in the different species of Burkholderia. The cellular fatty acids were converted to their respective fatty acid methyl esters (FAMEs). From the mass spectra obtained, it was obvious that only the carboxylic acid groups were derivatized. Saturated, unsaturated, hydroxyl and cyclo FAMEs from C4 to C18 were observed. The most common saturated fatty acids that were detected in the Burkholderia species include C12:0, C14:0, C15:0, C16:0, iso C17:0, C18:0 and C22:0. The most common hydroxy fatty acids that were detected include C14:0 2-OH, C14:0 3-OH, C16:0 2-OH, C16:0 3-OH and C18:0 10-OH. The unsaturated and cyclo fatty acids were C16:1ω7c, C18:1ω7c and cyclo C17:0.
The poly(3HBA-co-3HVA) polymers were fragmented into six important biomarkers. The biomarkers were identified as 3-butyric acid methyl ester (3-BAME), 2-butyric acid methyl ester (2-BAME), 3-valeric acid methyl ester (3-VAME), 2-valeric acid methyl ester(2-VAME), 3-hydroxy butyric acid methyl ester (3-HBAME) and 3-hydroxy valeric acid methyl ester (3-HVAME). The six peaks associated with these biomarkers are shown in the total ion chromatogram (TIC) of B. pseudomallei PHLS 72, which was cultured on blood agar at 37°C (Fig. 1). The source of these different peaks was confirmed by subjecting a poly(3HBA-co-3HVA) standard to the TCM procedure. The TIC of the poly(3HBA-co-3HVA) standard after the TCM procedure is shown in Fig. 2.
The different fragments that were generated by exposing the poly(3HBA-co-3HVA) standard to the TCM procedure revealed information regarding the reactions that occur during the procedure itself. Initially, the C-O bond between the acyl group and the oxygen was cleaved and hydrolysed to the parent carboxylic acids, 3-hydroxy butyric acid (3-HBA) and 3-hydroxy valeric acid (3-HVA). These carboxylic acids were then methylated into their corresponding methyl esters (3-HBAME and 3-HVAME). After dehydration and methylation, each of the acids yielded two additional compounds, 2-BAME and 3-BAME or 2-VAME and 3-VAME (Fig. 2). The biomarkers derived from 3-HBA (3-BAME, 2-BAME and 3-HBAME) were more abundant than the biomarkers from 3-HVA (3-VAME, 2-VAME and 3-HVAME). This resulted from the standard polymer containing more 3-HBA than 3-HVA.
Algorithm construction and testing
The biomarkers were tested to see whether they could be used to differentiate between the select agents and the other Burkholderia species. The C14:0 2-OH biomarker was found exclusively in B. pseudomallei, which allowed us to easily distinguish B. pseudomallei from the other species (Fig. 3). The six biomarkers that resulted from fragmenting the poly(3HBA-co-3HVA) polymers via the TCM procedure were present in all of the Burkholderia species. Nevertheless, these biomarkers were found to be useful in differentiating between B. mallei and the near neighbours. An unidentified unsaturated aldehyde peak and the ratio of C18:0 to C14:0 were also found to be useful in differentiating B. mallei from the near neighbours. The unidentified aldehyde and the six fragments of poly(3HBA-co-3HVA) were more abundant in the near neighbours than in B. mallei. Fig. 4 shows the extracted ion chromatogram (m/z 69) of 2-BAME and demonstrates the difference in the abundance of these biomarkers between B. mallei and the near neighbours.
The differentiation algorithm was used to predict the identities of all samples that were prepared in this study; a schematic of the final differentiation algorithm is shown in Fig. 3. The results from applying the algorithm to the samples are summarized in Table 2. Of the 25 samples that were misidentified, 13 were B. vietnamiensis, 11 were B. cepacia, and only one was B. multivorans. All of these samples were incorrectly identified as B. mallei because they displayed an unusually low abundance of the six biomarkers that were generated from poly(3HBA-co-3HVA). In addition, the 13 samples of B. vietnamiensis exhibited large differences in the abundance of important biomarkers. The variation in these samples made it difficult to effectively use the ratio of C18:0 to C14:0, which is typically higher in B. vietnamiensis, to identify these samples. However, this ratio was generally higher among B. vietnamiensis samples when compared to samples with roughly equal abundance. The peak area of each biomarker was divided by the peak area of C16:0 3-OH, which is a biomarker that displayed approximately equal relative abundance in all strains. This step was taken to make all biomarkers invariant to abundance. The standard deviations for quantification of amounts or ratios of biomarkers used are included in Table S1. Although there are some large standard deviations reported, the mean values of these measurements across the bacterial species are sufficiently separated to allow efficient differentiation (Table 2).
Table 2. Evaluation of the automated differentiation algorithm against samples from various Burkholderia species
While testing the algorithm against the various samples that are described above, we also investigated the detection limit of this new method. We determined that the limit of detection for B. thailandensis is approximately 4000 cells when using the 2-BAME biomarker. In actuality, the abundance of 2-BAME is lower in B. thailandensis than in B. pseudomallei. Therefore, we would expect the detection limit to be lower for B. pseudomallei.
MIDI detection method
As stated previously, the MIDI method was used to confirm the identities of the various Burkholderia species that were used in this study. This method correctly identified all of the B. pseudomallei and B. mallei strains. However, it was unable to correctly differentiate the near neighbours from B. pseudomallei and B. mallei. For example, B. thailandensis was incorrectly identified as Escherichia coli, Shigella sonnei, Shigella flexneri, B. pseudomallei and B. mallei. Similarly, B. cepacia and B. vietnamiensis were misidentified as B. gladioli and B. mallei. Additionally, the method was simply unable to find a match for B. multivorans.
In this study, we were able to differentiate B. pseudomallei from B. mallei and the near neighbours based on the presence of C14:0 2-OH. This biomarker was found exclusively in B. pseudomallei, which is in agreement with several published studies (Inglis et al. 2003; Levy et al. 2008; Novem et al. 2009). In fact, Novem et al. (2009) proposed that the C14:0 2-OH fatty acid in B. pseudomallei might allow the bacterium to evade immune responses and thus avoid being cleared from the host. In addition, they proved that the C14:0, C14:0 2-OH, C14:0 3-OH and C16:0 3-OH fatty acids are derived from lipid A that is the innermost region of the lipopolysaccharide molecule.
We investigated the possibility of using the C16:0 2-OH fatty acid to aid in the differentiation of B. pseudomallei from the other species of Burkholderia. It was previously reported that the C16:0 2-OH fatty acid was absent in B. thailandensis, but present in B. pseudomallei (Levy et al. 2008). However, we detected this biomarker in both of these species. Given this information, we chose not to use the C16:0 2-OH fatty acid in our differentiation assay.
The C12:0, C14:0 and C18:0 saturated fatty acids were some of the most stable (i.e. independent of growth conditions) biomarkers that were observed in this study. In 2006, Krejci and Kroppenstedt reported similar observations. In addition, they reported that B. vietnamiensis had a lower ratio of C14:0 to C18:0 than other species within the B. cepacia complex. They used this ratio to differentiate B. vietnamiensis from other members of the B. cepacia complex. We used the inverse of this ratio to assist in the differentiation of B. mallei from the near neighbours.
Burkholderia species are often considered promising candidates for the biosynthesis of substantial amounts of poly(3-HBA-co-3HVA). However, we observed that different amounts of poly(3HB-co-3HV) were formed by each Burkholderia species. For example, the near neighbours displayed very low abundances of poly(3HBA-co-3HVA), and B. mallei displayed even less. The low levels of expression in these species of Burkholderia required us to use the C18:0 to C14:0 ratio, an unidentified unsaturated aldehyde peak (m/z 55), and the six biomarkers that were generated from poly(3HBA-co-3HVA) to differentiate B. mallei from the near neighbours.
We are currently developing statistical methods to identify other biomarkers that might aid in differentiation; several biomarkers have already been identified. Preliminary results indicate that these biomarkers will allow us to greatly improve the differentiation of B. cepacia, B. multivorans and B. vietnamiensis from each other and from B. mallei. We are also investigating the ability of this method to detect and differentiate each of the Burkholderia species from environmental samples and mixed samples.
Despite the continuing research, the method described in this study represents a novel approach for the differentiation of B. pseudomallei, B. mallei, B. thailandensis and several members of the B. cepacia complex (B. cepacia, B. multivorans and B. vietnamiensis) using a GC-MS system. The correct identification of these microbes from the environment is essential to research involving disease transmission and epidemiology in endemic areas. The results of this study indicate that the method is fast, accurate and simple to use. In addition, the results show that the algorithm is robust against different growth conditions (medium and temperature). This assay may also prove beneficial in a clinical diagnostic setting, where the rapid identification of B. pseudomallei and B. mallei is essential to effective treatment. In addition, this method could be easily employed after a suspected biological attack to confirm the presence of either B. pseudomallei or B. mallei.
This study was supported by a grant from the Department of Homeland Security (Contract No. HSHQDC-10-C-00136).