Introduction

Burkholderia cepacia complex (Bcc) is a subgroup within Burkholderia genus formed by closely related species with high similarity of 16S RNA gene (>97.7%), large genomes (7–9 Mb), and moderate levels of DNA–DNA hybridization (30–60%) [1].

Currently, Bcc comprises 23 related species associated with poor prognosis and rapid decline in lung function in patients with cystic fibrosis (CF). Burkholderia multivorans and Burkholderia cenocepacia are the most prevalent and dangerous Bcc species among CF patients worldwide, associated with a more rapid decline in lung function and mortality after transplant [2, 3].

Among B. cenocepacia, two main lineages were delineated based on recA sequence similarity: IIIA and IIIB. The Bcc strains are intrinsically resistant to many antimicrobial agents and are virulent and highly transmissible among CF patients [4]. Chronic pulmonary infection could persist for years and might develop into fatal septic pneumonia (cepacia syndrome) [5]. It is important to highlight that pathogenicity within Bcc varies [2].

Due to a high level of similarity among Bcc species, accurate identification has been problematic, and most conventional and automated phenotypic tests have shown low accuracy. In addition, several strains belonging to Bcc have been identified as other gram-negative non-fermenters [1, 6].

Although widely used in bacterial systematics, studies have shown that 16S rDNA and hisA sequencing have limited discriminatory power to differentiate Bcc species. recA gene sequencing is currently considered the gold standard method for species identification but is not available in most clinical microbiology laboratories [7, 8]. Matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-ToF MS) has been considered a rapid, accurate and relatively inexpensive method for bacterial identification [9]. Many clinical laboratories have changed their workflow practices by adopting MALDI-ToF instruments [10].

Identifying Bcc to the species level is therefore critical since misidentification can lead to inadequate infection control precautions and therapeutic measures [1, 6]. In this context, we evaluated the performance of MALDI-ToF MS compared with recA sequencing using the recommended manufacturer guidelines score to identify Bcc species from CF patients. We additionally evaluated the identification performance at species level, considering the decrease in the cut-off score value.

Materials and Methods

Isolates

A total of 145 Bcc isolates were collected between 2010 and 2018 from 38 patients with CF in two reference centres in Rio de Janeiro, Brazil. The isolates were obtained from the following specimens: sputum (n = 111), oropharyngeal swabs (n = 26), bloodstream (n = 4), bronchoalveolar lavage (n = 3), and hepatic abscess (n = 1).

recA Gene Sequencing

For molecular identification (recA sequencing), bacterial DNA was extracted by the thermal lysis method [11]. The DNA concentrations of samples were quantified using Qubit® 2.0 Fluorometer (Invitrogen™, Eugene, OR, USA), and further submitted to polymerase chain reaction (PCR) amplification using the primers BCR1 (5′ → 3′TGAGCCGCCGCAAGAAGAA and BCR2 (5′ → 3′CTCTTCCATTTCGTCCTCCGC) [12].

We used 50 µL of PCR MIX, consisting of 0.2 U/µL Platinum Taq DNA polymerase, 1 µL (250 µM) each deoxynucleoside triphosphate -DNTP, 1.5 µL (1.5 µM) MgCl2, 5 × PCR buffer (5 µL), 20 pmol/μL of each primer, and PCR-grade water q.s. 100 µL. Amplification was performed using GeneAmp PCR System 9700 thermal cycler (Applied Biosystems®, California, USA), under the following conditions: 1 cycle at 95 °C for 5 min followed by 35 cycles at 95 °C for 45 s, 58 °C for 45 s, and 72 °C for 90 s. The final extension was performed using 1 cycle at 72 °C for 10 min.

The recA amplicons were purified with Wizard® SV Gel and PCR Clean-Up System (Promega®, Madison, WI, USA) and then sequenced using BigDye Terminator v.3.1 Cycle Sequencing Kit (Applied Biosystems®, Foster City, EUA), as recommended by the manufacturer. recA gene was sequenced using a combination of primers BCR1, BCR2, BCR3 (5′ → 3′GTCGCAGGCGCTGCGCAA), and BCR4 (5′ → 3′GCGCAGCGCCTGCGACAT) [12].

Geneious v. 7.1.5 (Biomatters®, Auckland, New Zealand) and Molecular Evolutionary Genetics Analysis (MEGA) v.7.0 software programs were used to assemble and analyze sequences, while the Basic Local Alignment Search Tool (BLAST®) v. 2.10.0+ (available at https://www.ncbi.nlm.nih.gov) was used to compare with Bcc-type strains sequences deposited in GenBank (NCBI, Bethesda, USA).

MALDI-ToF MS System

For the identification of Matrix-Assisted Laser Desorption/Ionization-Time-of-Flight Mass Spectrometry (MALDI-ToF MS) by the direct colony method, isolates were subcultured in trypticase soy agar with 5% sheep blood (SBA) (PlastLabor®, Rio de Janeiro, Brazil), at 35 °C ± 2 °C after 24 h incubation. A single colony from each sample was applied directly in duplicate on a stainless-steel plate, air dried, and overlaid by 1 μL of 98% formic acid and then applied 1 μL of matrix solution saturated of α-4-cyano-hydrocycinnamic acid.

Mass spectra (m/z) were acquired in a linear positive mode at laser frequency of 60 Hz within a range of 2000–20,000 Da by ionization of 240 shots, performed by the Bruker Daltonics Microflex LT instrument, using MALDI Biotyper software 3.1 (Bruker Daltonics, Leipzig, Germany), with a total of 8540 bacterial MSPs. Calibration was performed with the Bruker Bacterial Test Standard, while Burkholderia cepacia ATCC 25608, Pseudomonas aeruginosa ATCC 27853, and Escherichia coli ATCC 25922 were used as controls. Tests were performed in duplicates; lower scores were excluded in one replication.

MALDI-ToF MS spectra were analyzed using FlexAnalysis v. 3.3 (Bruker Daltonics GmbH, Leipzig, Germany). Species and genus identification were obtained when the scores were ≥2.0 and between ≥1.7 and <2.0, respectively [13].

Results

In this study, all 145 isolates were identified by recA sequencing as: B. vietnamiensis (n = 60; 41.37%), B. cenocepacia IIIA (n = 34; 23.44%), B. multivorans (n = 29; 20%), B. cenocepacia IIIB (n = 16; 11.03%), B. contaminans (n = 4; 2.75%), one B. cepacia (0.68%), and one B. stabilis (0.68%). Bruker MALDI-TOF Microflex System was able to identify 100% of Bcc isolates at the genus level. Using the score of log ≥ 2.00 for confident species level identification, 53.1% (n = 77) of isolates had identities that matched recA species. The major percentage of agreement was for B. vietnamiensis (78.3%) and for B. cenocepacia IIIA (53%). Interestingly, for B. cenocepacia III B, the percentages were lower (37.5%). Thirty-five (24.13%) isolates were misidentified; of these, 29 received a scores ≥2.00. Most of the misidentification occurred with B. multivorans (n = 20; 6.9%), which were recognized as B. vietnamiensis or B. cenocepacia (Table 1). There were no results in which species identification differed between the first and second columns of the Microflex’s report.

Table 1 Performance of MALDI-ToF MS and recA sequencing methodologies for identification of Bcc samples
Full size table

Samples correctly identified by Microflex System compared to recA sequencing were analyzed through the scatter plot, including the minimum, maximum, and average scores obtained by the species: 2.0514 for B. cenocepacia IIIA (1.722–2.424), 1.991 for B. cenocepacia IIIB (1.895–2.210), 2.054 for B. cenocepacia (only sample), 1.991 for B. multivorans (1.767–2.089), and 2.132 for B. vietnamiensis (1.549–2.435) (Fig. 1).

Fig. 1
figure 1

Scatter plot (minimum, maximum and average) scores obtained by MALDI-ToF MS identification of Bcc isolates

Full size image

By decreasing the cut-off score value to ≥1.70, results increased from 53 to 74.5% concordance at species level: from 78.33 to 90% for B. vietnamiensis, 52.04 to 85.29% for B. cenocepacia IIIA, 37.5 to 93.75% for B. cenocepacia IIIB, and 17.24 to 31.03% for B. multivorans.

Discussion

Bcc has a complex and dynamic taxonomic organization with new species identified or reclassified over the years [14]. The misidentification of these species represents a problem for the treatment of CF patients and for epidemiological purposes, becoming a challenge for microbiology laboratories [15].

The MALDI-ToF Microflex System has been used to identify Bcc recovered from CF patient samples presenting variable results for species level in agreement with recA sequencing [4, 10, 16].

Previous studies using the MALDI-ToF Microflex System for Bcc species identification show agreement rates with other methodologies ranging from 97 to 100% identification at genus level and from 23 to 97% of correct identification at species level [4, 15,16,17,18,19]. Our study corroborates the accuracy from the MALDI-TOF Microflex System for identification at genus level. However, it showed only a correlation of 53.1% at species level.

B. contaminans and B. multivorans clinical isolates were misidentified by the MALDI-ToF Microflex System, as also previously reported by Desai et al. [17] and Fehlberg et al. [10]. Our study showed a higher frequency of misidentification for B. contaminans isolates, and the single B. stabilis isolate was also misidentified. These misidentifications can be attributed to the high genetic similarity of the isolates with similar mass spectra, varying according to the databases, as well as with the absence of their spectrum in the database itself [18, 20].

For CF patients, the worst scenario is the misidentification of B. cenocepacia as being from other species, once this species is of particular concern due to its high transmissibility among patients and association with a fatal septic pneumonia (cepacia syndrome). Five B. cenocepacia isolates were misidentified by the MALDI-ToF Microflex System with score ≥2.00.

The variable results among studies may be partly due to the MALDI-ToF Microflex System used; the version of the spectral matching used by the database version; the high degree of genetic homology shared by Bcc isolates, and the number of different species included [17, 18].

Some studies show protocols adapted to optimize the accuracy of identification, especially for microorganisms with difficult characterization, such as a decrease in the cut-off point for identification at species level proposed by manufacturers.

Barberis et al. [21] validated a score of 1.70 for species level identification in a large and diverse collection of Gram-positive rods, and Alatoom et al. [22] evaluated identification of 92 clinical isolates of Corynebacterium species by the MALDI-ToF Microflex System compared to molecular identification (rpoB or 16S rRNA gene sequencing), showing that all 80 (87%) isolates were correctly identified at the species level, considering a score >1.70. So far, there are no studies with Bcc that evaluate the identification performance of the MALDI-ToF Microflex System by decreasing the species cut-off interpretation values.

As a limitation of our study, since 145 strains were recovered from 38 patients, some of them may consequently pertain to the same clone.

Conclusions

The MALDI-ToF Microflex System was able to identify 100% of Bcc isolates at the genus level. Accuracy declines for species level identification (53.1%), using the score of log ≥ 2.00 for confident species level identification. Allowing for the inference of a decreasing cut-off score value to ≥1.70, the results show an increment to 74.5% agreement at species level. A lower score for species identification level must be validated with a large collection of Bcc species. Epidemiological monitoring and correct identification of Bcc species for CF patients are extremely necessary through techniques with a high degree of precision, pace, and applicability in the clinical microbiology workflow, using a technique equivalent to a molecular reference method. The MALDI-ToF Microflex System has proven useful at the genus level identification; however, it still needs improvements that allow more precise identification, requiring continuous updates and addition of new spectra to its database.