Verification of a Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry Method for Diagnostic Identification of High-Consequence Bacterial Pathogens

Abstract

We examined the utility of a single matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry method for the identification of security-sensitive biological agents (risk group 3 bacterial pathogens). The goal was 2-fold: to verify a method for inclusion into our scope of accreditation, and to assess the biological safety of extractions. We developed our sample flow to include a tube-based chemical extraction, followed by filtration, before processing on MALDI-TOF MS instruments in a containment level 2 laboratory.

TEXT

Matrix-assisted laser desorption–ionization time of flight mass spectrometry (MALDI-TOF MS) is a rapid, robust, and cost-effective technology, which is emerging as a routine bacterial identification method in clinical diagnostic laboratories (1–3). Despite its utility, there are concerns with the safety of sample processing and the quality of spectral databases for the identification of security-sensitive biological agents (SSBAs). As our laboratory predominately works with SSBA bacterial pathogens (the risk group 3 [RG-3] pathogens: Francisella tularensis, Yersinia pestis, Bacillus anthracis, Brucella species, and Burkholderia pseudomallei), our goal was to verify the usefulness and safety of a MALDI-TOF MS method for inclusion into our scope of accreditation (ISO 17025). We chose to pursue a full-tube-based protein extraction method involving cellular disruption and filtration, rather than the direct-colony plate (smear) method (4, 5), in an effort to mitigate the risk of viable organisms.

Over a period of 1 year, 146 bacterial samples (including 57 SSBA samples; Table 1) were processed as previously described (6). The bacteria were previously identified by standard genotypic and phenotypic methods. In brief, all cultures were chemically extracted in a microcentrifuge tube using 70% ethanol, 70% formic acid, and acetonitrile, and all SSBA bacteria were extracted in a biosafety level 3 (BSL-3) laboratory. SSBA extracts from spore-forming bacteria were filtered through a 0.1 μM microcentrifuge filtration unit (EMD Millipore, Etobicoke, Ontario, Canada), as recommended by Dauphin and Bowen (7). The viability of the extract was performed by inoculating 10% of the extract onto both sheep's blood agar (SBA) (or cysteine heart agar [CHA] for F. tularensis) and into 2 ml of tryptic soy broth (TSB), and incubating at 35°C and 5% CO₂ for 48 h. After 48 h, the TSB was further subcultured to SBA or CHA and monitored for an additional 72 h. Any bacterial growth was tested with agent-specific assays to rule out the SSBA, indicating a contaminant. After a lack of growth was confirmed, extracts were removed from the BSL-3 laboratory, and 1.5 μl of the extract was applied (at a minimum of four replicate spots) to an MSP-96 MALDI target plate (Bruker Daltonics, East Milton, Ontario, Canada), with an overlay of 1.5 μl of α-cyano-4-hydroxycinnamic acid (CHCA) matrix solution (50% acetonitrile and 2.5% trifluoroacetic acid). Mass spectra were acquired with the FlexControl software (version 3.0) on a Bruker Daltonics microflex LT mass spectrometer, which was located in a BSL-2 laboratory area. The data were searched against proprietary Bruker Daltonics Biotyper Taxonomy (version 3.0; n = 5,637 mass spectral profiles [MSPs]) and Biotyper Security-Relevant (SR; n = 123 MSPs) reference library databases. A local custom-developed MSP database consisting of BSL-3 and BSL-2 MSPs (n = 92) was generated to supplement the proprietary databases. Triplicate extractions of each bacterial species culture were spotted 8-fold, for a total of 24 spots per custom MSP database entry. The Biotyper match scores for SSBA bacteria, which are presented in Table 1 for each database search, were recorded as the aggregate mean of each species (including standard deviation and range). We report here the primary database matches for the SSBA species tested and a secondary near-neighbor species match for comparison. The bacterial species were identified as defined by the Bruker Biotyper software as highly probable species identification (score, >2.3), secure genus identification, probable species identification (score, 2.0 to 2.29), probable genus identification (score, 1.7 to 1.99), and as no reliable identification (score <1.7).

TABLE 1.

Verification of the Bruker microflex LT MALDI-TOF MS (using Biotyper software) for the identification of bacterial cultures^a

Species	n	Biotyper matchb	Bruker database resultsc			Local database results
			Mean ± SD score	Score range	Species-level ID % (>2.0)d	Mean ± SD score	Score range	Species-level ID % (>2.0)d
Y. pestis	6	Y. pestis	2.141 ± 0.068	2.081–2.237	100	2.604 ± 0.035	2.403–2.782	100
		Y. pseudotuberculosis	2.122 ± 0.084	2.004–2.237	100	2.498 ± 0.051	2.444–2.540	100
F. tularensis	9	F. tularensis	2.012 ± 0.072	1.727–2.208	56	2.391 ± 0.060	2.156–2.698	100
		Francisella philomiragia	1.467 ± 0.081	1.373–1.583	0	1.556 ± 0.108	1.276–1.726	0
B. anthracis	7	B. anthracis	2.233 ± 0.064	2.145–2.345	100	2.299 ± 0.047	2.104–2.661	100
		Other B. cereus complex spp.e	2.023 ± 0.056	1.591–2.273	57	2.147 ± 0.056	1.520–2.511	57
B. pseudomallei	5	B. pseudomallei	2.054 ± 0.112	1.887–2.124	100	2.614 ± 0.102	2.420–2.722	100
		B. mallei	1.996 ± 0.107	1.764–2.139	40	NA	NA	NA
B. abortus	5	B. abortus	NA	NA	NA	2.681 ± 0.035	2.548–2.797	100
		Other Brucella spp.f	2.042 ± 0.073	1.785–2.303	100	2.274 ± 0.074	2.178–2.435	100
B. canis	2	B. canis	NA	NA	NA	2.585 ± 0.084	2.410–2.761	100
		Other Brucella spp.f	2.133 ± 0.072	2.016–2.251	100	2.373 ± 0.067	2.319–2.426	100
B. melitensis	12	B. melitensis	2.056 ± 0.062	1.748–2.279	67	2.499 ± 0.068	2.027–2.728	100
		Other Brucella spp.f	NA	NA	NA	2.292 ± 0.075	1.888–2.466	92
B. ovis	2	B. ovis	NA	NA	NA	2.780 ± 0.039	2.728–2.772	100
		Other Brucella spp.f	2.194 ± 0.043	2.177–2.212	100	2.372 ± 0.027	2.350–2.393	100
B. suis	9	B. suis	NA	NA	NA	2.546 ± 0.048	2.132–2.762	100
		Other Brucella spp.f	1.927 ± 0.081	1.772–2.182	100	2.331 ± 0.049	2.136–2.460	100

^aSSBAs (n = 57) were extracted and confirmed to be not viable in BSL-3 containment before testing on the Bruker microflex LT instrument. Only relevant SSBA bacterial species and selected near-neighbor species are included for comparison. NA, not applicable, as the Bruker reference library database only contains MSPs for B. melitensis, and the local database does not contain B. mallei.

^bThe Biotyper scores are aggregate means of scores from each extract, with minimum of 4 replicate spots for each individual culture (see Materials and Methods), and the standard deviation among replicates is presented.

^cBruker Daltonics Taxonomy (n = 5,687) and Security-Relevant SR (n = 123) reference library databases.

^dThe percent identification is the proportion of cultures tested achieving the Bruker Biotyper threshold for species-level identification (score, >2.0). The range is the range of mean Biotyper scores from each individual species.

^eIncludes B. cereus, B. thuringiensis, B. weihenstephanensis, and B. mycoides but not B. anthracis.

^fIncluding all Brucella species other than the sample culture being analyzed.

Notably, the tube-based extraction method (without a filtration step) in use within our BSL-3 laboratory did not inactivate all B. cereus complex species. Growth was observed following extraction in 3 of 31 (9.7%) Bacillus species samples (one B. anthracis and two B. thuringiensis), and incorporating a filtration step (filter pore size, 0.1 μm) to the extract yielded no growth upon viability testing. This is in agreement with Drevinek et al. (8), who suggested a filtration step for highly virulent pathogens. Thus, it is unlikely that a 70% formic acid extraction alone, or a direct colony extraction method incorporating 70% formic acid treatment, is a safe process for extracting SSBA pathogens for MALDI-TOF MS analysis. Cunningham and Patel (9) recently suggested that 70% formic acid treatment may be sufficient for safe analysis of potentially hazardous bacterial cells, including Gram-positive spore-forming bacteria from the B. cereus complex (BCC). However, their suggestion was based on a single B. cereus strain used as a surrogate for B. anthracis. Our data demonstrate that extraction methods without a filtration step do not inactivate all BCC bacterial extracts, but additional studies are needed to contribute to the development and definition of safe methods for the elimination of possible exposure to viable SSBA material within a laboratory setting. The method used here was simple and efficient but does not take into account other factors that can affect viability, such as standard inoculum loads or extended incubation times. For this reason, we recommend to err on the side of caution and confirm sample inactivation of SSBA material before testing with MALDI-TOF MS. Further, the growth of extracts without a filtration step demonstrates the utility of the viability protocol in use; the need to dilute out chemicals that might inhibit growth is of importance.

Our work demonstrated that the nonviable filtered Bacillus species extracts successfully yielded consistent Biotyper match scores, and the genetically homogeneous BCC species could not be confidently differentiated, as all had comparable Biotyper results against both the proprietary and local databases (Table 1). Although B. anthracis cultures had 100% species-level matches (score, >2.0), we observed preferential matches to other BCC species over B. anthracis in the Biotyper software within individual culture replicates, which is of concern for laboratories that utilize databases lacking B. anthracis spectra. There are reports of B. anthracis-specific biomarker peaks (10), but these peaks were not considered outside our routine Biotyper software analysis.

This is also true of the closely related Y. pestis and Yersinia pseudotuberculosis (11), and unsurprisingly, MALDI-TOF MS diagnostic tests have been reported to incorrectly identify Y. pseudotuberculosis as Y. pestis (12). In our study, they could not be differentiated at the species level with either database, as both cultures had high Biotyper match scores (>2.4) and equivalent 100% bacterial identification at the highly probable species identification level (Table 1). With the exception of Yersinia enterocolitica, the MALDI-TOF MS test could not discriminate species among the genus Yersinia, although the reported Yersinia-specific biomarkers (13) were not part of this study. Based on our data, diagnostic laboratories that do not have the Bruker Daltonics SR or custom database might incorrectly identify Y. pestis as Y. pseudotuberculosis at high Biotyper species-level identification scores (and vice versa), since only Y. pseudotuberculosis MSPs are included in the standard Bruker Taxonomy database.

Another homogeneous group, the Brucella spp., were also not differentiated at the species-level. B. melitensis, B. abortus, B. canis, B. suis, and B. ovis cultures all matched to the locally generated custom database spectra of multiple Brucella spp. at probable species identification scores of ≥2.0 (100% identification; Table 1). This confirms the results from Ferreira et al. (14), who reported reliable identification of Brucella spp. to the genus level only, although Lista et al. (15) were able to differentiate Brucella spp. with a custom reference database. Notably, the Biotyper SR database contains B. melitensis reference spectra only, while the standard Taxonomy database does not contain any Brucella species MSPs. We are currently working toward a significant expansion of our local high-quality Brucella MSP library for improved identification of diagnostic samples.

F. tularensis cultures (100%) were differentiated at the species level in the locally generated custom database, while the proprietary Bruker Biotyper database identified only 56% of the F. tularensis cultures at the species level. This result agrees with the result of Seibold et al. (16), who identified cultures of Francisella spp. to the species level with a supplemented Biotyper database. B. pseudomallei matched Biotyper scores at the species-level threshold (>2.0) based on the Bruker database, which was only slightly higher than that for the related SSBA bacterial species, Burkholderia mallei (Table 1). The local database produced higher match scores for B. pseudomallei, and we plan to expand the local MSP entries, as previous studies have recommended the use of an appropriate supplemental reference library for the genus Burkholderia (17).

The standard Bruker Daltonics Biotyper database does not contain SSBA pathogens, such as B. anthracis, which are only available through the separate SR library database, requiring special clearance for acquisition and use. Therefore, it is likely that SSBA cultures can be incorrectly, and unknowingly, identified by the software as a species-level match to a BSL-2 near-neighbor-species profile. The importance of using the Biotyper SR library has been described (18), and a comparison of experimental sample mass spectral profiles to a database supplemented with locally generated MSPs provides higher and more-confident match scores than those with the standard Biotyper database (Table 1) (6, 19, 20). In this method verification, we employed a locally generated custom MSP database, which contributes to the development of a high-quality curated database for high-consequence SSBA pathogens using the Bruker guidelines for library creation (21). Lasch et al. (10) recently described an international ring trial for MALDI-TOF MS identification of high-consequence pathogens, wherein identification accuracy improved with a special reference library. Thus, the complementation of the standard off-the-shelf database would assist in alerting a clinician that a possible SSBA organism may be present. Unfortunately, even with the use of an in-house supplemental database, we could not differentiate with confidence most SSBA pathogens to the species level. Consequently, this MALDI-TOF MS method gained accreditation only as a confirmatory method for identifying (or ruling out) SSBA bacteria at the genus-level only. This test result is not by itself a public health-actionable result for high-consequence pathogens.

In summary, for the MALDI-TOF MS identification of high-consequence SSBA bacteria we (i) used a full-tube-based chemical extraction (ethanol-formic acid-acetonitrile) with an additional postextraction filtration step, and (ii) conducted the identification of samples by comparison to MALDI-TOF MS databases supplemented with locally verified SSBA (RG3 select agent) spectra to trigger the recognition of the possibility of an SSBA being present.