Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry for the identification of Burkholderia pseudomallei from Asia and Australia and differentiation between Burkholderia species

Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is increasingly used for rapid bacterial identification. Studies of Burkholderia pseudomallei identification have involved small isolate numbers drawn from a restricted geographic region. There is a need to expand the reference database and evaluate B. pseudomallei from a wider geographic distribution that more fully captures the extensive genetic diversity of this species. Here, we describe the evaluation of over 650 isolates. Main spectral profiles (MSP) for 26 isolates of B. pseudomallei (N = 5) and other Burkholderia species (N = 21) were added to the Biotyper database. MALDI-TOF MS was then performed on 581 B. pseudomallei, 19 B. mallei, 6 B. thailandensis and 23 isolates representing a range of other bacterial species. B. pseudomallei originated from northeast and east Thailand (N = 524), Laos (N = 12), Cambodia (N = 14), Hong Kong (N = 4) and Australia (N = 27). All 581 B. pseudomallei were correctly identified, with 100% sensitivity and specificity. Accurate identification required a minimum inoculum of 5 x 107 CFU/ml, and identification could be performed on spiked blood cultures after 24 hours of incubation. Comparison between a dendrogram constructed from MALDI-TOF MS main spectrum profiles and a phylogenetic tree based on recA gene sequencing demonstrated that MALDI-TOF MS distinguished between B. pseudomallei and B. mallei, while the recA tree did not. MALDI-TOF MS is an accurate method for the identification of B. pseudomallei, and discriminates between this and other related Burkholderia species.


Introduction
The genus Burkholderia is composed of Gram-negative species that are predominantly nonpathogenic environmental saprophytes. A clinically important exception is Burkholderia pseudomallei, which causes an estimated 165,000 cases of melioidosis per year worldwide, of which 89,000 are predicted to be fatal [1]. A high proportion of reported cases occur in northeast Thailand, where melioidosis is the second most common cause of community-acquired bacteremia and the third most common cause of death from infectious diseases [2]. B. pseudomallei receives additional attention because of its biothreat potential [3], and because of infection in travelers returning from endemic regions [4,5]. In Thailand, B. pseudomallei is widely distributed in soil and environmental water [6][7][8] and may be found together with the closely related B. thailandensis, which is very rarely associated with human disease [9].
Making an accurate diagnosis of melioidosis is key to patient outcome since empiric antimicrobial therapy for sepsis does not include the first-line drugs recommended for melioidosis (ceftazidime or a carbapenem). Diagnosis relies on bacterial isolation and identification since clinical manifestations lack specificity. B. pseudomallei may be isolated from a range of clinical specimen types, but half of melioidosis cases have positive blood cultures. Direct identification of B. pseudomallei in positive blood cultures can reduce the time to diagnostic confirmation by 24 hours. This can be achieved using a rapid immunofluorescent assay (IFA), in which the most common antibody described is a monoclonal that recognizes B. pseudomallei capsular polysaccharide (CPS) [10]. This reagent is also used in a latex agglutination assay that can identify B. pseudomallei directly in blood cultures, other sample types, and colonies picked from culture media [11]. A monoclonal antibody against CPS has also been incorporated into a lateral flow assay [12], which can detect B. pseudomallei directly in a range of clinical samples and cultures.
The recent recognition of a B. thailandensis variant that expresses B. pseudomallei-like capsular polysaccharide (Bp-like CPS) [13] has important implications for the accuracy of B. pseudomallei identification assays based on antibody detection of CPS, since this gives a false positive result. The potential relevance of this has increased with the recent finding that B. thailandensis expressing Bp-like CPS is widely distributed in the environment in Thailand (V. Hantrakul, personal communication). Occasional cross-reacting isolates of B. cepacia have also been observed (D. Dance, personal communication). Methods that distinguish between B. pseudomallei and other Burkholderia species such as B. thailandensis include arabinose assimilation, more extended biochemical testing such as that incorporated in the commercial API 20NE biochemical kit (bioMérieux), real-time PCR, and sequencing [14][15][16][17][18][19][20]. Automated identification systems are available in many laboratories, although these may misidentify B. pseudomallei as B. cepacia [21][22][23][24].
Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been increasingly introduced into diagnostic laboratories for rapid bacterial identification. The first published study of its application to Burkholderia spp. evaluated 10 B. pseudomallei and 17 B. mallei isolates and generated reference spectra data [25]. Several case reports have since reported its use to diagnose melioidosis, including infection in returning travelers [5,[26][27][28][29]. It has also been reported to differentiate between B. pseudomallei wild-type and single gene mutants [30], to delineate clustering in a collection of 11 B. pseudomallei from northeast Thailand [31], and to determine ceftazidime resistance simultaneously with B. pseudomallei identification [32]. An inter-laboratory trial for the identification of highly pathogenic bacteria using MALDI-TOF MS included one isolate each of B. pseudomallei and B. mallei and concluded that a compilation of complete and comprehensive databases with spectra from a broad strain collection was of paramount importance for accurate microbial identification [33]. The largest evaluation of B. pseudomallei to date was undertaken using 66 B. pseudomallei. Nearly all of these (62/66 isolates) were initially misidentified as B. thailandensis using an existing database (DB5627), but an enhanced database subsequently identified all correctly [34]. A limitation of this study was that most of the isolates (63/66) were from a single country (Taiwan), the remaining 3 originating from Beijing, China.
At the time that this study was performed, representation of the 30 Burkholderia species in the biotyper database was limited and did not include B. pseudomallei, B. ubonensis, B. oklahomensis, B. humptydooensis or B. mallei. This indicates a need to expand the reference database and potentially increase the sensitivity and specificity of the assay. Furthermore, the need remains to undertake an evaluation of MALDI-TOF MS to identify B. pseudomallei drawn from a wider geographic distribution, which is important because of the genetic distinctiveness of isolates from Asia compared with Australasia and the highly plastic and variable genome, both of which could impact on the performance of MALDI-TOF MS. Previous studies have also lacked the inclusion of B. thailandensis expressing Bp-like CPS. The aim of this study was to expand the MALDI-TOF MS database with Burkholderia spp. and then evaluate this technology using a large collection of B. pseudomallei together with other Burkholderia spp. including B. thailandensis expressing Bp-like CPS. We added value to our findings by testing its performance using blood cultures spiked with B. pseudomallei, and evaluating the similarity between B. pseudomallei clusters arising from MALDI-TOF MS data compared with phylogenetic characterization based on recA sequence data.

Bacterial isolates
B. pseudomallei and other Burkholderia species used in this study are detailed in Table 1  pseudomallei was performed by the submitting laboratories using a range of different methods that reflected variation in clinical and research practice and the lack of a single gold standard method. In Asia, the predominant method used was a combination of colony morphology, antibiotic susceptibility pattern, arabinose assimilation and latex agglutination [35]. Elsewhere, colony appearance followed by API20NE (bioMérieux) was commonly used. Other Burkholderia species were identified using one or more of biochemical methods, including 16S rRNA sequence, recA sequence and DNA-DNA hybridization [14][15][16][17][18][19][20]. Additional, non-Burkholderia species were identified using standard laboratory methodology supplied by the Department of Medical Science, Ministry of Public Health, Thailand. All isolates were shipped in accordance with international guidelines to an accredited BSL-3 laboratory in Bangkok where the evaluation was conducted. Unless otherwise specified, bacteria were cultured on Columbia agar at 37˚C in air for 24 h. All isolates were stored in trypticase soy broth (TSB) with 15% glycerol at -80˚C.

Latex agglutination test and multilocus sequence typing
Latex agglutination reagent based on the 4B11 monoclonal antibody to capsular polysaccharide of B. pseudomallei was prepared and the assay performed as previously described [11,35]. Multilocus sequence type (ST) was known for 21 B. pseudomallei, 21 B. mallei and 10 B. thailandensis isolates, which have been reported previously [13,36,37].

MALDI-TOF MS analysis
Protein was extracted from bacteria using the formic acid extraction method as previously described [38,39], with several modifications. In brief, a loopful of bacteria was harvested from an agar plate and suspended in 300 μl ultra-pure distilled water (UDW), to which 900 μl absolute ethanol (Merck, Darmstadt, Germany) was added. The suspension was centrifuged at 16,200 g for 2 min and the pellet left to dry at room temperature. Twenty-five microliters of 70% formic acid (V/V) (Sigma-Aldrich, Fluka, MO, USA) was added and mixed thoroughly, followed by an equal volume of acetonitrile (Merck, Darmstadt, Germany). The mixture was centrifuged at 16,200 g for 2 min, and then 1 μl of supernatant was spotted onto a MSP-384 polished steel target plate (Bruker Daltonics, Germany). After drying in air, all spots were overlaid with 1 μl of matrix, α-cyano-4-hydroxycinnamic acid (HCCA) (Bruker Daltonics, Germany) dissolved in a solution of 50% acetonitrile, 2.5% trifluoroacetic acid and 47.5% water (Sigma-Aldrich, Fluka, MO, USA). Each spot was measured in 200 shot steps for a total of 1200 laser shots using an MALDI-TOF Mass Spectrometer Autoflex speed (Bruker Daltonics, Germany) and FlexControl software (version 3.4.135, Bruker Daltonics, Germany). Spectra were obtained in the linear positive mode with an accelerating voltage of 19.5 kV and analyzed within a mass range of 2,000-20,000 Da. Before measurement, the instrument was calibrated using the Bacterial Test Standard (BTS) following the manufacturer's instructions (Bruker Daltonics, Germany). E. coli DH5-Alpha was used as a positive control and matrix solution was used as a negative control. Identification was achieved using the MALDI-Biotyper software (version 3.1, Bruker Daltonics, Germany). These experiments were performed in duplicate and the scores were averaged. Interpretation was performed according to the manufacturer's recommendation; score of ! 2.3, reliable species level identification; 2.0-2.29, probable species level identification; 1.7-1.9, probable genus level identification; 1.7, unreliable identification [40]. In a pilot study, at least 4 colonies (size~1 mm) were required for B. pseudomallei to be identified with a score of ! 2.3. H2708a, ST60; 1106a, ST70; and 1026b, ST102). These STs also represent the major B. pseudomallei STs in Thailand [37]. B. mallei evolved from a single lineage of B. pseudomallei and the two isolates selected here (NCTC 10247 and Mongolia 1) were used previously as reference isolates is a study that described the discrimination between B. mallei and B. pseudomallei [25]. B. ubonensis, B. oklahomensis, B. humptydooensis and B. vietnamiensis are rarely described and all isolates belonging to these species in our collection were used. B. cepacia, B. multivorans and B. thailandensis were added since they may be misidentified as B. pseudomallei by some identification systems.

MALDI-TOF MS reference database
Identification of specific peaks for the discrimination of Burkholderia species The Clinprotools software (version 3, Bruker Daltonics, Germany) was used to identify peaks that discriminated between Burkholderia species. This was performed using 20 replicates of a single protein preparation from each of nine Three independent experiments (20 replicates for each experiment) were performed and the results of each experiment checked for consistency of the discriminating peaks.
The statistical algorithms incorporated in the Clinprotools software (Quick Classifier (QC)/ Different Average, Supervised Neural Network (SNN) and the Genetic Algorithm (GA)) were used to analyze protein peaks between 2,000 and 20,000 Da. Potentially discriminatory peaks were identified based on high reliable prediction and separation based on cross validation values (>90%), as described previously [30]. Three statistical tests (Anderson-Darling (AD), ttest/ANOVA (TTA), and Wilcoxon/Kruskal-Wallis (W/KW) are incorporated into Clinprotools and were used to analyse intensity data. All discriminatory peaks that exhibited P-values of <0.01 by any of these statistical test were further evaluated by eye [41] and validated with the peaks of other isolates from the same species. Peaks that were unique to specific species or statistically different in intensity and showed consistent results from three independent experiments were considered as the discriminatory peaks. All Burkholderia isolates used the for identification of discriminating peaks were used for the construction of a dendrogram.

MALDI-TOF MS dendrogram
To construct the dendrogram, Flexanalysis software (version 3.4) was used to adjust the baseline and smoothness of the spectra. Twenty spots of a single protein extract for each isolate were used to construct the main spectrum profiles (MSP) using the MALDI-Biotyper software (version 3.1). The following parameters were used: the Biotyper MSP creation standard method was used, with a maximum mass error of each single spectrum of 2000, desired mass error for the MSP of 200, desired peak frequency minimum of 25%, and maximum desired peak number for the MSP of 70.  H2660a, H2708a, 1106a and 1026b) and a further 16 isolates each assigned to a different ST (ST696, ST76, ST77, ST80, ST345, ST40, ST126, ST177, ST70, ST501, ST102, ST10, ST54 and ST60 (Fig 1). B. thailandensis and B. thailandensis with BP-like CPS capsule were randomly selected from our freezer archive. We included all of the B. mallei (N = 21), B. humptydooensis MSMB43 (N = 1), B. oklahomensis (N = 3), B. cepacia (N = 8), B. vietnamiensis (N = 1), B. ubonensis (N = 1) and B. multivorans (N = 1) in our collection. Cluster analysis was performed based on comparison of MSP using the following setting parameters: the distance measure was set to Spearman, the linkage was set to single, and score threshold value for a single organism was set at 1000.

Determination of minimum bacterial input for accuracy of MALDI-TOF
The minimum number of bacteria in suspension and minimum number of colonies required to achieve an accurate MALDI-TOF result were determined. To quantify the number of bacteria, the experiment was performed using bacterial suspension. B. pseudomallei K96243 was harvested from an overnight culture of Columbia agar, suspended in sterile saline, adjusted to approximately 1 x 10 7 CFU/ml and then serially diluted from 1 x 10 7 CFU/ml to 10 CFU/ml. Bacterial cells in 1ml of each dilution were harvested by centrifugation at 16,200 g for 2 min. Protein was extracted from each pellet prior to MALDI-TOF analysis as above. The bacterial count was confirmed using Columbia agar spread plates in triplicate and colony counts after overnight incubation at 37˚C. To determine the minimum number of colonies required for MALDI-TOF, between 1 and 10 colonies (size~1 mm) of B. pseudomallei K96243 were harvested from Columbia agar plates using a loop, suspended in UDW and the protein extracted from each. Extracts were analyzed in triplicate in both assays. Spectra with maximal absolute peak intensities ranging from 10 3 to 10 4 arbitrary units were considered for evaluation [42,43].

Effect of culture media
The effect of culture media was examined for five B. pseudomallei isolates (K96243, NR9921, 1106a, 1026b and 576a). Bacteria were streaked onto Ashdown agar, blood agar, Columbia agar, chocolate agar, Luria-Bertani (LB) agar, MacConkey agar, Mueller-Hinton agar, trypticase soy agar and incubated at 37˚C in air for 24 hours and harvested using a 1 μl loop (! 10 colonies, size~1 mm). Bacteria were suspended in 300 μl UDW and the protein extracted and analyzed in triplicate using MALDI-TOF MS.

MALDI-TOF MS identification of B. pseudomallei in spiked blood culture
Three BACTEC Plus Aerobic/F bottles were each inoculated with 10 ml of whole blood drawn from a single healthy volunteer. An overnight culture of B. pseudomallei K96243 was adjusted to approximately 1 x 10 8 CFU/ml, serially diluted to a concentration of 100 CFU/ml, and 100 μl (10 CFU) inoculated into each bottle. The inoculum was confirmed using colony counts on agar plates. Bottles were incubated at 37˚C with 200 rpm shaking. At 12, 16, 20 and 24 h after incubation, one ml from each bottle was withdrawn for protein extraction using ammonium 0.826% NH 4 Cl (W/V) as a lysis buffer, as described previously [44,45] with modifications. Briefly, an equal volume of 0.826% NH 4 Cl was added to 1 ml of blood culture fluid, mixed and centrifuged at 16,200 g for 2 min. The pellet was lysed twice as above before washing twice with 1 ml ultrapure distilled water and centrifuged as before. The supernatant was discarded and the pellet used for protein extraction.
RecA sequencing recA sequencing and phylogenetic analysis was performed as previously described [20].  1) and B. multivorans (N = 1). Genomic DNA was extracted and PCR amplification performed using BUR3, BUR4 and BUR5 primers as previously described [20]. Products were visualized using agarose-gel electrophoresis and purified using ExoSAP-IT PCR Product Cleanup (Affymetrix UK Ltd., UK). Purified PCR products were sequenced by Macrogen Inc. (Korea). Nucleotide sequence of B. thailandensis E555 was obtained from the NCBI database (accession no. AECN01000010.1). All sequences were aligned and trimmed to a 348 bp region using Clustal W using MEGA software version 7 [46]. A maximum likelihood tree was constructed using the Nearest-Neighbor-Interchange (NNI) and Tamura-Nei model [47] using MEGA software version 7.0.14 [46].

Ethical approval
The study was approved by Ethics Committee of the Faculty of Tropical Medicine, Mahidol University (approval number MUTM 2016-034-01). The principal investigator's blood was

Evaluation of MALDI-TOF MS for the identification of B. pseudomallei
Twenty replicates of B. pseudomallei K96243 were tested using MALDI-TOF MS and the Bruker Daltonics database version 3.3.1.0. These were all identified as B. thailandensis, with a median score below that for reliable species identification (median 1.96, range, 1.83-2.04). We noted that the Biotyper database did not contain B. pseudomallei, B. mallei, B. ubonensis (Table 2). B. pseudomallei isolates showed some variability in similarity scores against the 5 B. pseudomallei added here to the reference profiles. Despite this, the highest score was to B. pseudomallei for all 581 B. pseudomallei tested, with the second and third highest score being B. mallei (median score 2.21, range 1.68-2.46, IQR 2.14-2.29) and B. thailandensis (median score 2.01, range 1.62-2.22, IQR 1.95-2.06). The specificity of MALDI-TOF MS was evaluated by testing 25 isolates including B. mallei (N = 19), B. thailandensis (N = 6), and 23 isolates belonging to 21 species in other Genera (see Methods). These bacterial species were correctly identified, indicating 100% specificity.

Effect of culture media on MALDI-TOF MS identification
We evaluated whether the culture medium used to grow bacteria prior to MALDI-TOF MS affected the accuracy of identification. Five B. pseudomallei strains (NR9921, 1106a, 576a, K96243, 1026b) were grown overnight on eight different solid media. All isolate/media combinations were identified as B. pseudomallei, with an identification score of at least 2. 45

Identification of B. pseudomallei in spiked blood culture
Since the speed of identification of B. pseudomallei from blood culture is crucial for patient management, we tested whether MALDI-TOF MS would give an accurate identification of B. pseudomallei K96243 in spiked blood culture fluid. For this we used three 30 ml BACTEC Plus Aerobic/F blood culture bottles each containing 10ml of blood from a healthy volunteer, spiked with 10 CFU, and tested after 8, 10, 12 and 24 h of incubation at 37˚C with shaking. Testing at 8, 10, 12 h showed unidentified or misidentified results, but accurate identification for B. pseudomallei was obtained at the 24 h time point (median identification score 2.25 (range 2.17-2.36)). Colony counting demonstrated that the bacterial concentration in the blood culture bottle fluid at 24 h was 2.1 x 10 9 CFU/ml.

Latex agglutination test
The latex agglutination test is used to rapidly identify B. pseudomallei colonies in our laboratory and elsewhere, and so we performed a comparative assessment

Identification of discriminatory peaks
Twenty discriminatory peaks were identified from MSP spectra of representative isolates from 9 Burkholderia species and B. thailandensis with Bp-like CPS with cross validation results ranging between 98% and 100% (Fig 2 and S1    Phylogeny based on MALDI-TOF MS protein profiles and recA A phylogenetic tree based on the MSP of 67 Burkholderia isolates divided nine species into two major branches (Fig 1).  [36,37] and are shown in Fig 1. B. mallei isolates belonged to a single ST (ST40), while B. pseudomallei belonged to 8 different STs. A phylogenetic tree based on recA sequence was compared with the MALDI-TOF dendrogram for 26 isolates representing nine Burkholderia species (Fig 3). This demonstrated a broadly similar structure between the two, with distribution of the species between two major divisions. A notable difference was that MS distinguished between B. pseudomallei and B. mallei, while the recA tree did not.

Discussion
MALDI-TOF MS is increasingly used for bacterial identification in diagnostic microbiology laboratories. This technology can reduce time of identification since bacteria are sampled directly from bacterial colonies and the test takes around 30 minutes to perform. The rapid microbiological diagnosis of melioidosis is essential because clinical features are non-specific, empiric drug regimens are sub-optimal, and infection is often fatal without appropriate treatment. This is compounded by the fact that healthcare providers in non-endemic countries may not readily recognize melioidosis. MALDI-TOF MS could also be used for environmental surveys looking at the distribution and presence of B. pseudomallei and related species.
Our findings confirm that MALDI-TOF MS can accurately identify B. pseudomallei regardless of geographic origin, and was able to reliably distinguish between this, other Burkholderia spp. and a range of other common pathogens, provided that the database is suitably modified. Since the top hit identification for each tested B. pseudomallei isolates could be any of five B. pseudomallei isolates in our reference database, we recommended laboratories that wish to identify B. pseudomallei using this method to extend the database by adding reference profiles of all five B. pseudomallei isolates. Diagnostic laboratories use a range of culture media, but we demonstrated that this has no effect on the performance of the test. We also showed that MALDI-TOF MS accurately identifies B. pseudomallei in blood cultures, which could reduce the time taken to diagnostic confirmation and appropriate antimicrobial treatment. Our results confirm a previous report in which MALDI-TOF was used to identify B. pseudomallei from blood cultures from two septicemic patients in Australia [29]. The simulated blood culture experiment in which 10 CFU was used as the starting inoculum reflects the bacterial load in blood during human infection, which has been reported previously to be a median count of 1.1 CFU/ml [48]. Our observation that identification of B. pseudomallei by MALDI-TOF was not accurate until after an incubation period of 24 hours implies that in clinical practice, this method will be effective on bottles that have flagged in an automated incubation system or after 24 hours or more of incubation.
The Clinprotools software used to support the interpretation of MALDI-TOF MS identifies species-specific peaks as the basis for species identification [30,41,49]. In agreement with previous studies [25,30], we identified specific peaks at 4,410, 5,794, 6,551, 7,553 and 9,713 Da for all B. pseudomallei isolates tested. We also observed specific peaks for B. ubonensis (2,908 Da), B. vietnamiensis (3,932 Da) and B. thailandensis variants expressing Bp-like CPS (5,835 Da). The peak at 2,049 Da may be used to differentiate between B. pseudomallei and B. mallei because this was present in all B. pseudomallei isolates but not in any B. mallei isolates. We also noted several peaks which displayed significantly higher peak intensities in specific species, which may be useful for the discrimination of B. humptydooensis (2,049 Da), B. multivorans Latex agglutination is a sensitive screening test for suspected B. pseudomallei, but positive latex agglutination results have been described previously for B. mallei, B. thailandensis with Bp-like CPS, S. aureus [11,35,50] and some strains of B. cepacia (unpublished data), potentially leading to confusion in diagnostic laboratories if used alone. This study shows that MAL-DI-TOF MS can be used to reliably distinguish between these organisms. Furthermore, both MSP and recA dendrograms confirmed that B. pseudomallei, B. mallei and B. thailandensis were arranged in the same phylogenetic group.
In conclusion, MALDI-TOF MS is an accurate and discriminatory tool for the identification of B. pseudomallei if sufficient MSPs are added in the Biotyper database. MALDI-TOF MS could be used to increase the rapid detection of cases of melioidosis in the clinical setting, and reduce time to appropriate antimicrobial therapy.
Supporting information S1