Ethics statement

In all cases, patients/volunteers or their legally authorized representative provided written informed consent and granted permission to store samples for future studies. The study was approved by the ethics committee of Hainan Medical University (HYLL-2022–265).

Main reagents

B. pseudomallei strain HNBP001, an epidemic strain isolated from a melioidosis patient with pneumonia in Hainan, China, was isolated and sequenced by our team. The genome sequence has been uploaded to the National Center for Biotechnology Information (NCBI). It is similar to those of previously published genome annotations of B.pseudomallei [17]. E. coli BL21 (DE3) cells were purchased from Zhuangmeng International Biology Co., Ltd. (Beijing China). The Pet-28a recombinant plasmid transformation was constructed by Liu Chengli from our laboratory and gifted with preferential treatment. The Genome Extraction Kit was purchased from Tiangen Co., Ltd. (Beijing China). The PCR product purification and plasmid extraction kits were purchased from Omega Bio-Tek (Norcross, GA, USA). The Prime-STAR Max-DNA Polymerase was purchased from TaKaRa (Tokyo, Japan). Restrictive endonucleases (Nde I, Xho I, and Hind III) and T4 DNA ligases were purchased from New England Biolabs (NEB, Beverly, MA, USA). Yeast and tryptone were purchased from OX-OID (Hampshire, UK). 50 × TAE Buffer, Sodium chloride was purchased from Xilong Science Co., Ltd. (Guangzhou, China). Goat anti-human IgM mu-chain (horseradish peroxidase) and Rabbit anti-goat IgM were purchased from Bio-Rad (Shanghai, China).

Study design

Wild-type Bp, HNBP001, was used as the infection bacterium to construct the infection model and extract EVs from the cell culture supernatant (solution 1). Meanwhile, HNBP001 was cultured separately in Luria-Bertani (LB) liquid culture medium overnight until an optical density at 600 nm (OD₆₀₀) of 1.0 was achieved. The supernatant of bacterial culture (solution 2) was obtained. EVs were separated by ultracentrifugation (S3 Fig) from the infection model (S1 and S2 Figs) and Bp. The EV proteins were hydrolyzed into peptide segments by proteases before mass spectrometry analysis. Subsequently, the measured peptide sequences were compared with known protein libraries for overall protein sequence analysis, ultimately obtaining information on the types and contents of proteins on each group of EVs. The study flowchart is shown in the Scheme 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Scheme 1. Study flowchart.

https://doi.org/10.1371/journal.pntd.0013543.g001

BEAS-2B cell culture

The cells were grown in DMEM containing 10% fetal bovine serum (FBS, EV-free), 100 mg/ml penicillin and 100 mg/ml streptomycin. The cells were cultured in the incubator for 1–2 days until the cells were completely covered with the bottom of the Petri dish. Then wash the cells twice with 1 × PBS buffer. The adhesion cells were eluted with trypsin, and transferred to a 15 ml sterile centrifuge tube to centrifuge the supernatant, and then the cells were resuspended with 1–2 ml medium. Take 300 µL cell suspension and add it into a new Petri dish containing 10 ml of culture medium and shake it up. Put the Petri dish into the CO₂ incubator at 37°C for culture.

Bacterial culture

Initially, remove the bacterial strain stored at -80°C and allow it to thaw at room temperature. Inoculate 10 µL of the bacteria onto solid LB medium and cultivate until the formation of a distinct single colony. Following this, select a single colony and inoculate it into liquid LB medium, placing it on a constant temperature shaker set at 37°C for overnight incubation at 200 rpm. Subsequently, transfer 1 mL of the bacterial culture into a 300 mL liquid LB culture flask, cultivating it until reaching OD₆₀₀ = 1.0 (designated as solution 1). The prepared LB liquid culture medium was filtered through a 0.22 µm membrane and sterilized under high pressure to eliminate yeast-derived particles and minimize the influence of vesicles produced by other bacteria on the results.

Building of infection model

The schematic diagram of infection model construction is shown in S1B Fig. Firstly, the BEAS-2B cell line was cultured until the cells covered the bottom of the culture dish, and cell counts were performed using a Neubauer hemocytometer counting chamber. Approximately 1 × 10⁵ cells were transferred to a 12-well plate and cultured overnight to allow the cells to fully adhere to the surface for growth. Diluted Bp bacterial suspensions were added to the cell culture dish at ratio of Bp: BEAS-2B = 1; 5; 10; respectively. The bacterial infection of cells lasted for 4, 6, and 9 hours, respectively. After infection, the culture medium was discarded, and the cells were washed twice with PBS. Subsequently, 5 mL of DMEM containing 750 ng/mL of Kanamycin and 750 ng/mL of Amikacin was added to the culture dish for 4 hours to kill all extracellular bacteria. The cells were then rinsed 2–3 times with PBS. Bacterial-infected cells were incubated in a new, antibiotics-free DMEM for 18–24 hours. The morphology of the cells, the number of intracellular bacteria, and cell activity were continuously observed. Finally, the supernatant was retained for subsequent extraction of EVs (solution 2).

Isolation of EVs

The extraction of EVs was based on previously reported methods for extracting EVs with slightly change [18] (S3 Fig). (1) The collected solutions (solution 1 and solution 2) were centrifuged at 300 g for 10 min at 4°C. (2) The supernatant was transferred to a labeled new centrifuge tube and centrifuged at 2,000 g for 10 min at 4°C to move the dead cells. Then centrifuged at 10,000 g for 20 min at 4°C again to move the cell debris. (3) A filter with a pore size of 0.22 µm was utilized, and collected the filtrate. (4) The collected supernatant from step 3 was transferred to a high-speed centrifuge tube, centrifuged at 150,000 g for 70 min at 4°C, and the supernatant was discarded. (5) The precipitate was resuspended with 2 mL of sterile PBS and then the volume was replenished to 10 mL. The sample was centrifuged at 150,000 g for 70 min at 4°C and the supernatant was discarded, resulting in the corresponding EVs. Supplementing phosphate buffered saline (PBS) was used to resuspend EVs and stored at −20°C, and extracted protein within one week.

Preliminary verification of protein expression differences by SDS-PAGE

1. 1). Sample preparation: based on the determined protein concentrations, extract an equal amount from each sample and transfer it into a centrifuge tube. Add an appropriate volume of sample loading buffer and heat the mixture to 100°C for 5 minutes to induce protein denaturation.
2. 2). Sample loading: carefully dispense 7 μ L of the protein marker and 20 μ L of the protein sample into the gel wells, followed by adding 10 μ L of 1 × loading buffer to seal the empty wells.
3. 3). Electrophoresis: set the voltage to 90 V to concentrate the proteins into a visible band. Once the marker band is distinct in the separating gel, adjust the voltage to 110 V to continue the electrophoresis until the proteins have migrated to the bottom of the gel.
4. 4). Silver staining: remove the gel from the apparatus and treat it with the silver staining reagent according to the specified protocol for protein visualization.

Mass spectrometric analysis of EVs protein

Mass spectrometry analysis

The dissolved peptides were separated using liquid chromatography on the Easy NLC 1200 Ultra Performance Liquid Chromatography system. Mobile phase A consisted of an aqueous solution with 0.1% formic acid and 2% acetonitrile, while mobile phase B contained 0.1% formic acid and 90% acetonitrile. The liquid phase gradient was set as follows: 0–68 min, 5–22% B; 68–82 min, 22–34% B; 82–86 min, 34–80% B; 86–90 min, 80% B. The flow rate was maintained at 500 μL/min. The separated peptide fractions were then analyzed using the Orbitrap Exploris 480 mass spectrometer with the NSI (nanospray ionization) ion source. The ion source voltage was set to 2300 V, and the FAIMS (high-field asymmetric waveform ion mobility spectrometry) compensation voltages were set to -45 V and -65 V. The parent ions of the peptides and their secondary fragments were detected and analyzed using high-resolution Orbitrap technology. The primary mass spectrometry scanning range was set from 400 to 1200 m/z, with a scanning resolution of 60,000. For secondary mass spectrum scanning, the fixed starting point was set at 110 m/z, with a resolution of 15,000. TurboTMT was turned off. Data acquisition used a data-dependent scanning (DDA) program, where, after the initial scan, the 25 peptide segments with the highest signal intensity were selected. Each selected segment underwent fragmentation in the HCD (higher-energy collisional dissociation) collision cell using 27% fragmentation energy. Second-order mass spectrometry analysis was performed sequentially for these fragmented segments. To optimize mass spectrometry utilization, the automatic gain control (AGC) was set to 100%, ensuring efficient ionization. A signal threshold of 50,000 ions/s was established. The maximum injection time was set to auto, adapting to the requirements of each injection. In addition, a dynamic exclusion time of 20 s was implemented during tandem mass spectrometry scanning to prevent repetitive scanning of parent ions. This dynamic exclusion strategy helps to enhance data quality and reduce redundancy in subsequent analyses.

Statistical analysis

The secondary mass spectrometry data was retrieved using a proteome discoverer (v2.4.1.15). The resulting peptide sequences were searched using the following parameters. The database used was Burkholderia_pseudomallei_strain_K96243_272560_PR_20220708.fasta, which contains 5,717 sequences. A decoy database was incorporated to calculate the false discovery rate (FDR) caused by random matches. Additionally, a common contaminants database was included to eliminate the impact of contaminant proteins on the identification results. The relative quantitative values of each protein in the comparative group samples were subjected to a t-test, and the corresponding P-value (P < 0.05) was calculated to indicate significance. A significant upregulation is defined as a change in differential expression levels exceeding 1.5, while a significant downregulation is defined as a change less than 2/3. This study selected bacterial-origin proteins from post-infection EVs for comparative analysis with pure bacterial EV proteomes, based on the following rationale: EV-enriched bacterial proteins demonstrate significantly enhanced recognition by the host immune system and must be Bp-specific (BLASTp-verified with less than 25% similarity to phylogenetically related species). Quantitative analysis revealed that bacterial-origin proteins constituted 19.9% of the total proteins in infected group EVs, with 27.9% of these being upregulated proteins. Biologically, infected cell EV proteomes primarily reflect host immune responses (83.6% host-derived proteins, such as inflammatory factors) and pathogen-host interface characteristics (16.4% immunodominant bacterial proteins). These host response patterns are common across various infections. Methodologically, direct comparison with pathogen-derived proteins filters out nonspecific host-response proteins, improves target identification specificity, and enhances diagnostic biomarker discovery efficiency by 41%, compared to conventional methods.

Candidate protein selection

This study established stringent protein screening criteria: candidate proteins must be exclusively derived from Bp components and exhibit ≥2.5-fold higher expression in EVs from infected host cells compared to Bp monoculture EVs. Additionally, the proteins were required to meet the following subcellular localization characteristics: (1) membrane localization (validated by PSORTb 3.0 prediction), (2) presence of ≥1 transmembrane domain, and (3) surface exposure with predicted immunogenicity.

Protein expression and purification

A protein expression vector was constructed through enzyme digestion and enzyme linkage techniques, based on the genome information of the HNBP001 strain, the expression primers were designed using oligo 7 software (S1 Table). The protein was expressed in E. coli BL21 cells (Novagen) nurtured in LB broth with 100 mg L ⁻¹ Ampicillin and induced by 1 mM IPTG. Before purification, E. coli cells were stored on ice in 20 mL of a nondenaturing lysis buffer. An ultrasonic crusher was used for 10–15 s to crush cells for 15 min until the cell suspension was transparent and centrifuged at 12,000 rpm for 20 min at 4°C. The clarified lysate was loaded onto a 2 mL Ni-NTA (nitrosamine triacetin acid) agarose column and eluted with a 50 mM imidazole buffer at pH 8 using 250 nM imidazole over 2 column volumes. Subsequently, the size and purity of the protein were verified using WB.

Serum samples

Serum samples from 43 melioidosis patients were collected at the Second Affiliated Hospital of Hainan Medical University between December 2022 and June 2023, with all cases confirmed by blood culture. Samples were obtained during the acute phase of illness, processed immediately for serum separation, and stored at -80°C until assay. Forty-seven serum samples from volunteers in non-endemic areas were collected from matriculating undergraduates at Hainan Medical University during routine entrance physical examinations. Volunteers were required to have no history of residence or travel in endemic regions or abroad before enrollment. Endemic regions were defined according to the 2022 Chinese Expert Consensus on Diagnosis and Treatment of Melioidosis, published in the Chinese Journal of Infectious Diseases. The consensus designates high-prevalence areas as Hainan Province, Guangdong Province, Guangxi Zhuang Autonomous Region, Fujian Province, Hong Kong Special Administrative Region, and Taiwan Province (https://doi.org/10.3760/cma.j.cn311365-20211113-00398).

21 cases serum samples from patients with single bacterial infections (such as E. coli, K. pneumoniae, S. maltophilia, A. baumannii, and P. aeruginosa, as shown in Table S5) were used to detect the specificity of biomarkers. Which recruited between December 2022 and June 2023 at the Second Affiliated Hospital of Hainan Medical University. The naturally acquired volunteers included asymptomatic volunteers from endemic areas (47 cases), laboratory partners (high-exposure individuals, 14 cases), clinically cured patients (5 cases), and individuals with occupational animal exposure in endemic regions, including naturally occurring high-risk individuals (farm workers, 11 cases) from a farm infected with Bp.

Serological validation

This study employed an IgM-specific ELISA, which was found to be sensitive and effective for the serological diagnosis of acute melioidosis [19]. An immune ELISA plate was prepared by immobilizing the expressed and purified protein to detect serum IgM antibodies. First, the expressed and purified protein is immobilized onto the wells of a 96-well plate, with approximately 250 ng of purified protein added to each well. The plate is incubated at 4°C overnight to allow the protein to bind to the plate surface. Unbound proteins are removed by washing the wells with PBS (phosphate-buffered saline). Next, any unoccupied gaps on the solid phase carrier surface are sealed using WBS (sealing buffer) at room temperature for 4 hours. Next, a serum sample diluted 1:50 with physiological saline is added to each well of the prepared immune plate and incubated at 37°C for 1 hour. This facilitates specific binding between serum antibodies and the immobilized protein. Then, excess serum is gently washed off 2–3 times using a PBS buffer containing 0.02% tween-20 to remove any unbound antibodies. The plate is then dried gently on absorbent paper. Next, sheep anti-human enzyme-linked secondary antibody is added to each well and incubated at 37°C for 30 minutes. This secondary antibody specifically binds to the human antibodies in the serum that have bound to the immobilized protein. The samples are then rinsed 2–3 times using PBS buffer containing 0.02% Tween-20 before adding ELISA chromogenic reagents. The results are observed within 1 min using a multifunctional ELISA reader. The observed color change from colorless to blue indicates a positive result, confirming the presence of the target protein in the serum sample.

Seropositivity was determined using a validated cut-off value defined as the mean optical density (OD₄₅₀ₙₘ) of negative controls (non-endemic healthy volunteers; n= ±3SD, : average OD₄₅₀ of IgM-ELISA measurements representing non-endemic healthy volunteers, SD: Standard deviation). Samples exceeding this threshold were considered positive. Diagnostic performance was evaluated by calculating sensitivity, specificity, positive/negative predictive values (PPV/NPV), and area under the curve (AUC) using culture/PCR-confirmed melioidosis as the reference standard.

Extraction of EV and biomarker selection

EV was isolated via ultracentrifugation (S3 Fig) from both the infection model and Bp cultures (S1 and S2 Figs). As shown in transmission electron microscopy (TEM) and the NTA (S4 Fig), EV from individual Bp (Fig 1Aa, 130 ± 10 nm)/ BEAS-2B (Fig 1Ab, 172 ± 10 nm) and infection cell (Bp/BEAS-2B, Fig 1Ac, 166 ± 10 nm) exhibited a vesicular-like structure (Fig 1A). These findings are consistent with prior research on EV morphology [20]. The protein concentrations extracted from these EVs were sufficient for further experimental analyses (S1 Table). As shown in the SDS-PAGE (Fig 1B), there were significant differences between EV proteins from the infection cells and those from Bp, especially proteins of about 25, 55, and 70–100 kDa. The preliminary experimental results laid a solid foundation for subsequent protein analysis of EV.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. (A) TEM of EV from Bp (a), BEAS-2B (b), and Bp/BEAS-2Bs (c). (B) SDS-PAGE was used to analyze the protein expression in EV of each group. Mass spectrometry data quality control and sample repeatability testing: (C) peptide number distribution, (D) protein coverage distribution, (E) protein molecular weight distribution, (F) PCC, (G) PCA of all bacterial-origin proteins on EVs in the Bp monoculture group and the Bp/BEAS-2B group.

https://doi.org/10.1371/journal.pntd.0013543.g002

Mass spectrometry data quality control and sample repeatability tests were shown in Fig 1. An analysis of peptide segments identified in the mass spectrometry data revealed that most proteins were covered by 1–8 unique peptide sequences (Fig 1C). Most peptide segments are composed of 7–20 amino acids (S5 Fig), which conform to the general rules of protein determination based on enzymatic hydrolysis and mass spectrometry fragmentation methods. Notably, this analysis showed that most proteins had a coverage rate of less than 30% (Fig 1D). The molecular weights of the identified proteins were evenly distributed, mainly concentrated between 10–90 kDa (Fig 1E).

Our analysis revealed high consistency in protein profiles of EV within each group, as assessed by pairwise Pearson correlation coefficients (PCC) (Fig 1F). This finding was further supported by principal component analysis (PCA) of the quantitative proteomic data, where samples within each group clustered more closely than those from different groups (Fig 1G). In summary, we successfully analyzed the proteins of EV derived from various models, including individual bacterial strains and Bp/BEAS-2B co-cultures. Notably, the three control samples within each group displayed high consistency, supporting the reliability of our mass spectrometry data for further protein quantification.

In this study, a comprehensive proteomic analysis was conducted. A total of 613,554 mapping points were recorded, resulting in 56,596 effective maps and the identification of 1,380 peptide segments. From these data, a total of 625 proteins were identified, and 212 of these were quantified (S6 Fig). These proteins were selected based on the protein library of K96243. Subsequent bioinformatics analysis revealed significant differential expression between the Bp and Bp/BEAS-2B groups. A total of 154 proteins were differentially expressed between the Bp and Bp/BEAS-2B groups, with 111 proteins down-regulated and 43 up-regulated in the infected group (Figs 2A, 2B and S10). Among the 43 highly expressed proteins, four are involved in the biological formation of cell walls, membranes, or envelopes. Three are involved in intracellular transport, secretion, and vesicle transport, six in energy production and conversion, and fourteen in the transport and metabolism of amino acids, enzymes, nucleic acids, and carbohydrates (Fig 2D). Of these, 21 proteins displayed a fold difference of less than 3.0 between the infection and control groups, including four ribosomal proteins that typically function within cells (Table 1). Additionally, some proteins are specifically localized within the cell, such as single-stranded DNA binding proteins, ATP synthase subunits, glucose-6-phosphate 1-dehydrogenase, adenosylhomocysteinase, and putative cytochromes. Furthermore, lipoproteins serve as structural proteins for vesicles and exhibit low specificity in their binding interactions. We also found that seven proteins were localized to the outer membrane, while thirteen proteins were localized to the cell envelope (Fig 2C). The subcellular localization of some proteins remains undetermined. Cluster analysis revealed that proteins with an FC > 2 were generally components of the transport enzyme complex, which plays a key role in regulating the immune response of bacteria to hosts and their response to various pressures (Fig 2E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List the proteins with more than 3-fold difference in expression between Bp and Bp/BEAS-2B EV.

https://doi.org/10.1371/journal.pntd.0013543.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. (A) Distribution of proteins with different expression level and (B) volcano plots of EV proteomes in Bp and infection groups. The top 20 significantly upregulated (red) or downregulated (black) protein are highlighted. Classification of the identified proteins according to Cellular Component (CC) and (C) Clusters of Orthologous Groups (COG) of proteins (D). (E) cluster analysis results of EV proteomes in Bp and Bp/BEAS-2B (Cluster the proteins based on FC, Q1 (FC < 0.5), Q2 (FC: 0.5-0.667), Q3 (FC: 1.5-2.0), Q4 (FC > 2.0); Red: high enrichment level, blue: low enrichment level).

https://doi.org/10.1371/journal.pntd.0013543.g003

Previous research has suggested that outer membrane proteins, bacterial exotoxins, and proteins with a clear transmembrane structure may serve as potential biomarkers or vaccine targets [21]. The putative outer membrane copper receptor (POMCR) is a receptor protein that is located on the outer membrane of cells and is involved in copper ion transmembrane transport. It showed a notable increase in expression in Bp/BEAS-2B’s EV proteins—3.77 times higher than in Bp EV. A peptide segment identified from the Bp/BEAS-2B cell model EV perfectly matched the POMCR sequence, highlighting its specificity (Fig 3A). When this sequence was compared with all Burkholderia spp. proteomes in NCBI (including B. mallei, B. cepacia complex), and prevalent Gram-negative pathogens in endemic areas (e.g., Pseudomonas aeruginosa, Acinetobacter baumannii), It was predominantly found in 59 Burkholderia strains carrying the POMCR protein, of which 58 were B.pseudomallei. (S3 Table). The predicted structure, function, and protein-like properties of POMCR were retrieved from UniProt (https://www.uniprot.org/uniprotkb/Q63JH4/entry). The protein structure 100% consistent with this protein comes from two different Bp strains, while 34 other Burkholderia strains express similar proteins with 90% structural similarity. Twenty-three different Bp strains express this protein, suggesting its conservation within Bp (S2 Table). Moreover, the protein has a clear transmembrane structure (Fig 3Cii).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Mass spectrum of peptide segments and sequence specific analysis of peptide segments of POMCR (A, B) and PPEP (C, D).

(E) Structure prediction of selected proteins PPEP (i), POMCR (ii), BLF1 (iii), and omp A (iv).

https://doi.org/10.1371/journal.pntd.0013543.g004

Proteomic analysis revealed significant upregulation of the putative phage-encoded peptidoglycan-binding protein (PPEP) in the EVs from infected cell models, with a 10.65-fold increase in expression. This was confirmed by matching a peptide segment from the EV to the PPEP protein sequence (Fig 3C). Among 56 Burkholderia strains carrying the PPEP protein, 47 belonged to the Bp complex, including 37 Bp strains (78.7%), 4 B. thailandensis strains, 2 B. humptydooensis strains, and 2 B. oklahomensis strains. No strains of B. cepacia or Pseudomonas aeruginosa were detected when the peptide segment was compared using the NCBI database (S3 Table). The PPEP has peptidoglycan-binding sites (positions 10–64 aa) and N-acetyldeaminase activity (positions 90–262 aa) (Fig 3Ci, structure of PPEP), with 100% similarity to the peptidoglycan-binding protein of PK23 in Burkholderia (S2 Table). The putative peptidoglycan-binding domain, found at the N or C terminus of various enzymes involved in bacterial cell wall degradation, may serve a general peptidoglycan-binding function [22].

The protein was mainly found in Burkholderia phage and Bp. As the growth of bacteriophages and bacteria is a complementary and co-evolutionary process, this indirectly suggests that the protein plays a role in bacterial defense against bacteriophages. However, the subcellular localization and function of this protein remain unclear (https://www.uniprot.org/uniprotkb/Q63YM4/entry). An analysis of bacterial species with similar proteins reveals that the two proteins, including the previously discussed POMCR, are predominantly found in the genus Burkholderia, with limited similarity to proteins from other Gram-negative bacteria. This specificity highlights their potential as highly conserved diagnostic markers for Burkholderia species. Based on these findings, POMCR and PPEP proteins were selected as candidate antigens for melioidosis in the present study.

Preparation of IgM-ELISA Kit and serological verification

Due to the lack of commercially available antibodies for the candidate proteins, this study employed custom ELISA kits to assess the presence of specific IgM antibodies in serum samples. Firstly, a recombinant protein expression vector was constructed with C-terminal and N-terminal His-tags through enzyme digestion and linkage (S7 Fig). The recombinant protein was then purified and validated using Western blot (WB) analysis (S7C Fig). Subsequently, the purified proteins were used as antigens to detect antibodies in the serum of melioidosis patients and control samples under optimized conditions (S8 and S9 Figs, the details shown in supplementary materials). The antibody levels of the four EV-proteins showed significant differences between the patient and control groups (Volunteers from non-endemic areas, 47cases) (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Corresponding antibodies in serum of the patients and the control group were detected by anti-POMCR (A), anti-PPEP (B), anti-BLF1 (C), and anti-omp A (D) IgM-ELISA kits.

(****: P < 0.01).

https://doi.org/10.1371/journal.pntd.0013543.g005

Receiver operating characteristic (ROC, Fig 5) curves were plotted to determine the optimal diagnostic threshold for the IgM-ELISA of different proteins (Fig 4). For the POMCR-IgM-ELISA, setting the cutoff at OD₄₅₀ = 0.118 yielded an area under the curve (AUC) of 0.9872, with sensitivities of 93.02% and specificity of 97.92%, respectively. The 95% confidence interval ranged from 0.9713 to 1.000 (Fig 5A). However, one volunteer in the control group, who tested positive for POMCR antibodies, was later identified as coming from Guangdong, a known epidemic area for melioidosis. As a result, their data was excluded from the control group analysis to ensure accurate assessment of the test’s specificity. Comparative analysis with anti-PPEP (Fig 5B), anti-BLF1 (Fig 5C), and anti-omp A (Fig 5D) revealed CUT-OFF values of 0.1095, 0.125, and 0.1085, with sensitivities of 93.33, 86.05, and 90.7%, and AUC values of 0.969, 0.9621, and 0.976, respectively. The dashed line in Fig 4 demarcates these thresholds, with samples below this line classified as negative. Control group analysis showed minimal false positives: 1 for POMCR-IgM-ELISA, 2 for BLF1-IgM-ELISA, and 1 for ompA-IgM-ELISA, underscoring the diagnostic reliability of each individual protein.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. ROC curves of anti-POMCR (A), anti-PPEP (B), anti-BLF1 (C), and anti-omp A (D) based on the results of Fig 4.

https://doi.org/10.1371/journal.pntd.0013543.g006

Serological evaluation of melioidosis patients (Table 2) identified two cases negative for all four biomarkers, two cases negative for a single marker (either anti-PPEP or anti-BLF1), and two cases positive for one marker (either anti-PPEP or anti-POMCR). To optimize diagnostic accuracy, we established a tiered criterion for diagnosis: Confirmed melioidosis (+): requires positivity for two of three biomarkers (anti-BLF1 + anti-POMCR/ anti-PPEP). Melioidosis exclusion (-): applies to isolated anti-BLF1 positivity. Further testing is mandated for sole anti-POMCR or anti-PPEP positivity. Using these criteria, two patients were definitively ruled as negative, while two others exhibited ambiguous serological profiles. This protocol enhances detection specificity in resource-limited settings while reducing the risk of overdiagnosis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of four melioidosis patients in the prospective screening cohort.

https://doi.org/10.1371/journal.pntd.0013543.t002

Biomarker specificity

To verify the specificity of the candidate protein in diagnosing melioidosis, serum samples from patients with a single bacterial infection (E. coli, K. pneumoniae, S. maltophilia, A. baumannii, and P. aeruginosa) were collected (S5 Table). Fig 6A shows a clear distinction between the serum samples from melioidosis patients and those with other bacterial infections, using the four protein immunoassay kits. Among them, anti-POMCR, anti-PPEP, and anti-omp A each yielded one positive result, while anti-BLF 1 had two positive results. Notably, all four positive results originated from a patient infected by P. aeruginosa, underscoring the high specificity of our biomarkers in distinguishing Bp from other Gram-negative bacteria. Based on previous results from melioidosis patients, non-melioidosis patients, and the healthy control group, the sensitivities of anti-POMCR, anti-PPEP, anti-BLF1, and anti-ompA were 95.35%, 93.02%, 90.70%, and 88.37%, respectively. The detection accuracy of the four IgM-ELISAs was 96.4%, 96.4%, 92.8%, and 93.7%, respectively (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Sensitivity, specificity, and diagnostic accuracy of the POMCR, PPEP, BLF1, and omp A IgM-ELISAs.

https://doi.org/10.1371/journal.pntd.0013543.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. (A) The specificity of recombinant protein IgM-ELISAs in diagnosis melioidosis. POMCR (a), PPEP (b), BLF1 (c), and omp A (d). (B) Proteins IgM antibodies of POMCR (i), PPEP (ii), BLF 1 (iii), and omp A (iv) were employed as biomarkers for detecting serum from different populations.

https://doi.org/10.1371/journal.pntd.0013543.g007

Through serological testing in different populations, we found significant differences in the serum test results between melioidosis patients and other groups. No positive results were observed in the two high-risk populations (Laboratory partners and Farm labourers, Fig 6B). However, among the examinees in the province, one serum sample tested positive for all four biomarkers, while only the POMCR-IgM-ELISA results were positive in another sample. Additionally, one case in the BLF1 test showed a positive result in the serum from farm laborers. According to the diagnostic criteria we established, this result is considered a false positive. This finding underscores the need for continuous refinement and validation of our diagnostic approach to ensure specificity and reduce false positives, especially in diverse or non-endemic populations.