Evaluation of the Vitek MS v3.0 Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry System for Identification of Mycobacterium and Nocardia Species

ABSTRACT

This multicenter study was designed to assess the accuracy and reproducibility of the Vitek MS v3.0 matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry system for identification of Mycobacterium and Nocardia species compared to DNA sequencing. A total of 963 clinical isolates representing 51 taxa were evaluated. In all, 663 isolates were correctly identified to the species level (69%), with another 231 (24%) correctly identified to the complex or group level. Fifty-five isolates (6%) could not be identified despite repeat testing. All of the tuberculous mycobacteria (45/45; 100%) and most of the nontuberculous mycobacteria (569/606; 94%) were correctly identified at least to the group or complex level. However, not all species or subspecies within the M. tuberculosis, M. abscessus, and M. avium complexes and within the M. fortuitum and M. mucogenicum groups could be differentiated. Among the 312 Nocardia isolates tested, 236 (76%) were correctly identified to the species level, with an additional 44 (14%) correctly identified to the complex level. Species within the N. nova and N. transvalensis complexes could not always be differentiated. Eleven percent of the isolates (103/963) underwent repeat testing in order to get a final result. Identification of a representative set of Mycobacterium and Nocardia species was highly reproducible, with 297 of 300 (99%) replicates correctly identified using multiple kit lots, instruments, analysts, and sites. These findings demonstrate that the system is robust and has utility for the routine identification of mycobacteria and Nocardia in clinical practice.

INTRODUCTION

Mycobacterium and Nocardia are diverse groups of bacteria increasingly implicated in disease, and yet these pathogens are challenging to identify to the species level for most clinical laboratories (1–6). The Mycobacterium tuberculosis complex is a major public health concern that is responsible for significant morbidity and mortality worldwide. The World Health Organization (WHO) recently estimated that as many as 10.4 million new infections and 1.4 million deaths occurred worldwide due to M. tuberculosis complex in 2015 (3). Additionally, nontuberculous mycobacteria (NTM) (3–5, 7–12) and Nocardia (13–15) are important causes of invasive disease, especially in immunocompromised hosts. Our ability to identify these organisms to the species level is important given that susceptibility testing is not widely performed, except in the case of M. tuberculosis complex. Thus, physicians often must make decisions regarding therapy on the basis of identification alone (9, 10).

Over the past 50 years, identification of Mycobacterium and Nocardia has transitioned from phenotypic characterization to mycolic acid analysis using gas-liquid chromatography (GLC) or high-performance liquid chromatography (HPLC) and, most recently, to the use of gene-specific probes or DNA sequencing. The advent of molecular methods has enabled greater taxonomic differentiation of these organisms in a clinically meaningful time frame (7, 10). However, it is well recognized that targeting of a single gene is frequently incapable of discriminating between closely related species within these genera. For example, 16S rRNA gene sequences are highly conserved among mycobacteria. For the Nocardia, identification is complicated by the presence of multiple copies of the 16S rRNA gene (13, 15). For these reasons, multilocus sequence analysis using 16S rRNA, rpoB, erm, hsp65, gyrB, and/or secA1 genes has been recommended. Unfortunately, the cost and complexity of multilocus sequencing often restrict its use to reference laboratories, thus delaying identification of isolates from hospital-based clinical laboratories.

The application of matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) for the identification of Mycobacterium and Nocardia began over a decade ago, but significant variability in test methodologies and performance has been observed (16–40). The direct-spotting method typically used for Gram-negative and Gram-positive bacteria and yeast is not adequate for Nocardia and Mycobacterium species due to the hardiness of their cell wall; rather, mechanical disruption and protein extraction are necessary. These methods can be performed in the presence of ethanol to simultaneously inactivate infectious organisms (22). Recent research has also highlighted the need for well-curated and validated databases that include a sufficient number of isolates of each species to account for the inherent variability in spectra (18, 20, 25, 26, 39, 40). The objective of this multicenter study was to establish the accuracy and reproducibility of the standardized Vitek MS v3.0 system (Vitek) for the identification of commonly encountered mycobacteria and Nocardia in the clinical laboratory. This most recent update to the FDA-cleared Vitek system includes 19 Mycobacterium and 12 Nocardia species that could be expected to be commonly recovered from cultures of “acid-fast bacilli” on solid medium in the clinical laboratory (41).

MATERIALS AND METHODS

Study sites.

The performance of the Vitek MS v3.0 system was evaluated at four clinical laboratories within the United States, namely, LabCorp (Burlington, NC), ARUP Laboratories (Salt Lake City, UT), Memorial Sloan Kettering Cancer Center (New York, NY), and the University of Washington Medical Center (Seattle, WA). This study was approved by the human subjects committees at the respective sites, when deemed necessary by their institutional review boards.

Organisms.

Each study site prospectively collected and tested clinical isolates until they reached a minimum of 10 isolates per species of a predefined list of organisms with the goal of inclusion of at least 30 isolates per species from all sites combined. In the event that a site was unable to obtain 10 unique fresh isolates of a given organism, additional frozen stocks obtained from a culture collection or provided by the sponsor were permitted. In addition to the clinical isolates, a challenge set of 50 well-characterized organisms and a reproducibility panel consisting of five organisms were provided to each laboratory for testing.

Mycobacteria were cultured on one of the following solid media: Coletsos agar, Lowenstein-Jensen agar (L-J), Middlebrook 7H10 agar (7H10), or Middlebrook 7H11 agar (7H11). The cultures were incubated at 35 to 37°C with CO₂ for either 3 to 7 days (rapidly growing mycobacteria [RGM]) or 7 to 28 days (slow growers). The exceptions were M. haemophilum, which was preferentially grown on Lowenstein-Jensen agar with 2.5% ferric ammonium citrate at 35 to 37°C with CO₂ for 7 to 28 days or at 25 to 33°C with CO₂ for 7 to 21 days; and M. xenopi, which was incubated at 35 to 37°C with CO₂ for 7 to 28 days or at 42°C with CO₂ for 7 to 28 days. Nocardia were cultured on buffered charcoal yeast extract agar (BCYE), chocolate agar (CHOC), Columbia blood agar (CBA), or Sabouraud dextrose agar (SDA) under aerobic conditions at 35 to 37°C for 18 to 72 h. Isolates were tested when sufficient growth was observed. The age of the culture and type of medium used were documented for each isolate. No differences were detected in the identification of organisms grown on different medium types or incubated for various periods of time. This is likely because these were the same culture conditions as were used to develop the v3.0 knowledge base.

Sample preparation.

Isolates were prepared for analysis using a Mycobacterium/Nocardia reagent kit (bioMérieux, Durham, NC) according to the manufacturer's instructions. This included mechanical disruption using 0.5-mm-diameter glass beads and bead beating for 5 min or vortex mixing for 15 min; the two methods have been shown to be equally effective at inactivating pathogens (42). This was followed by a 10-min incubation in 70% ethanol at room temperature and then protein extraction using 70% formic acid and acetonitrile. One microliter of the protein extract was then applied to a single spot on the target slide (Vitek MS-DS; bioMérieux), allowed to dry, and overlaid with 1 μl of α-cyano-4-hydroxycinnamic acid (Vitek MS-CHCA matrix; bioMérieux). Target slides were kept at room temperature and analyzed using Vitek MS v3.0 within 72 h of preparation.

Calibration and quality control.

For instrument calibration, an Escherichia coli reference strain (ATCC 8739) was transferred to designated wells on the target slide using a 1-μl loop, overlaid with 1 μl of Vitek MS-CHCA matrix, and air dried. Positive-control organisms (N. farcinica, ATCC 3308, and M. smegmatis, ATCC 19420) were inactivated and extracted using the same protocol as that described above and analyzed alongside a negative control (reagents alone) on each day of testing at each site.

MALDI-TOF analysis for organism identification.

The Vitek MS v3.0 system includes an OEM (original equipment manufacturer)-labeled Shimadzu Axima Assurance mass spectrometer linked to a reference database. Spectra are subjected to proprietary processes called “mass binning” and “normalization” and were then queried against the reference database to determine the taxonomic identity of the unknown organism, as previously described (43). Spectra were acquired and analyzed using Myla 4.0 software with database v3.0.

Reference method for organism identification.

Reference identification for all study isolates was performed at the University of Texas Health Science Center (Tyler, Texas). Partial 16S rRNA genes and several additional housekeeping genes (the rpoB [beta subunit of RNA polymerase], hsp65 [65-kDa heat shock protein], secA [ATPase component of bacterial secretion pathway], internal transcribed spacer [ITS], gyrB [B subunit of DNA gyrase], and erm [erythromycin ribosomal methylase] genes) were sequenced to identify the isolates. Isolates that could not be resolved to the species or complex level using these gene targets were additionally subjected to 16S rRNA complete gene sequencing. Partial 16S rRNA gene sequencing was performed on the samples using MicroSeq rDNA PCR and sequencing kits according to the recommendations of the manufacturer (Life Technologies, Carlsbad, CA). The resulting ∼500-bp gene was analyzed using RipSeq software (Pathogenomix, Santa Cruz, CA). Complete 16S rRNA gene sequencing was performed as previously described by Edwards et al. (44) using slight modifications. Sequencing of region V of the rpoB gene (720 bp, excluding the primer regions) was performed on the isolates utilizing primers MycoF and MycoR as described by Adékambi et al. (45). erm(41) gene sequencing was performed on M. abscessus complex isolates as described previously (46). Nocardia isolates were identified by sequencing of a 468-bp fragment of the secA1 gene as previously described by Conville et al. (47). The 156-amino-acid deduced SecA1 protein sequences were compared to an in-house database of 50 Nocardia type strain sequences. Sequences were compared to validated type strains and to all available sequences using RipSeq. Interpretation was in accordance with the Clinical and Laboratory Standards Institute (CLSI) interpretive criteria for DNA target sequencing (48).

Analysis.

The Vitek result was considered accurate to the species level if a single identification was given and if it matched the identification obtained by the reference method. It was considered correct to the complex or group level if a single identification or multiple identifications within the same genus were reported and if these results were within the same complex or group as the species identified using the reference method. It was considered incorrect if a single identification was given that did not match the result obtained by sequencing at the species, complex, or group level. In those instances in which no identification was obtained, repeat testing of a single spot was performed using the same extract. If this first repeat failed to result in an identification, a new sample extract was prepared and tested on a single spot. If an identification was still not made, a result of “no identification” was used as the final result. If the identification was made after respotting or spotting a new extract, then it was used as the final result.

Reproducibility testing.

Reproducibility testing was performed by two operators at each of three sites. A panel of five organisms (M. abscessus, M. chelonae, M. smegmatis, N. wallacei, and N. otitidiscaviarum) was tested in duplicate on two runs daily for 5 days. The identity of each organism was blinded with respect to the operators. Testing was performed using three different lots of reagents. The position of each organism on the target slide was predetermined, and the organisms were tested sequentially on one slide and in a randomized manner on a second slide. Sample preparation, organism identification on the Vitek MS v3.0, and result analysis were performed as described above.

RESULTS

Overall performance for clinical isolates.

A total of 963 clinical isolates, including 651 Mycobacterium and 312 Nocardia isolates, were tested. For the Mycobacterium isolates, 614 (94%) were identified with a single correct identification to the species, group, or complex level (Table 1). Of the remaining isolates, four (<1%) were misidentified and 33 (5%) could not be identified despite repeat testing. For the Nocardia isolates, 236 (76%) were correctly identified to the species level, with an additional 44 (14%) identified at the complex level (Table 2). Ten (3%) Nocardia isolates were misidentified, and 22 (7%) were not identified despite repeat testing. Overall, the Vitek system provided a single correct identification to the species, group, or complex level for the majority (894 of 963, 93%) of the Mycobacterium and Nocardia isolates tested.

TABLE 1.

Accuracy of identification of mycobacteria^a

Reference identification	No. (%) of isolates with indicated Vitek MS v3.0 ID result
	Correct to species level	Correct to complex or group level	Incorrect	No ID
M. tuberculosis complex (n = 45) (includes but does not discriminate among M. africanum, M. bovis/BCG, M. canettii, M. microti, M. pinnipedii, and M. tuberculosis		45 (100)
NTM slow growers
M. avium complex
M. avium (n = 68)	68 (100)
M. intracellulare (n = 43)	42 (98)	1 (2)
M. chimaeraN (n = 16)		16 (100)
MAC-XN (n = 20) (includes but may not discriminate among M. arosiense, M. bouchedurhonense, M. colombiense, M. marseillense, M. timonense, M. vulneris, and M. yongonense)		14 (70)		6 (30)
M. simiae complex
M. simiae (n = 28)	28 (100)
M. lentiflavum (n = 37)	33 (89)			4 (10)
M. shigaenseN (n = 3)				3 (100)
M. gordonae (n = 30)	28 (93)			2 (7)
M. gordonae/paragordonaeN (n = 3)		1 (33)	1 (33)	1 (33)
M. haemophilum (n = 18)	17 (94)			1 (6)
M. kansasii (n = 31)	31 (100)
M. malmoense (n = 4)	4 (100)
M. marinum (n = 25)	25 (100)
M. nebraskense* (n = 1)	1 (100)
M. paraenseN (n = 1)				1 (100)
M. parascrofulaceumN (n = 2)			1 (50)	1 (50)
M. paraffinicumN (n = 9)			2 (22)	7 (78)
M. scrofulaceum (n = 15)	15 (100)
M. szulgai (n = 24) Does not discriminate from M. angelicumb	23 (96)			1 (4)
M. xenopi (n = 24)	24 (100)
Total (n = 402)	339 (84)	32 (8)	4 (1)	27 (7)
NTM rapid growers
M. chelonae/abscessus group
M. abscessus complex (n = 40); includes but does not differentiate subspecies abscessus, massiliense, and bolletii		40 (100)
M. chelonaec (n = 29)	28 (97)			1 (3)
M. immunogenum (n = 19)	19 (100)
M. fortuitum-smegmatis-farcinogenes group
M. fortuitum group (n = 49); includes but does not discriminate between M. alvei, M. brisbanense, M. boenickei, M. farcinogenes,d M. fortuitum, M. houstonense, M. neworleansense, M. peregrinum, M. porcinum, M. senegalense, M. septicum, M. setense		48 (98)		1 (2)
M. smegmatise (n = 21)	21 (100)
M. mageritense* (n = 1)f	1 (100)
M. mucogenicum group
M. mucogenicum (n = 20)g	18 (90)			2 (10)
M. phocaicumN (n = 21)		20 (95)		1 (5)
M. mucogenicum/phocaicumh (n = 1)		1 (100)
M. neoaurum* (n = 1)	1 (100)
Total (n = 202)	88 (44)	109 (54)		5 (2)
Mycobacterium species (n = 2)		1 (50)		1 (50)
Total (n = 651)	427 (66)	187 (29)	4 (<1)	33 (5)

^aID, identification; ^N, species not included in v3.0 database; *, organism included in v3.0 database but not included in FDA claim.

^bM. angelicum has the same first 500-bp rRNA gene sequence as M. szulgai; M. angelicum was not recognized in the gene sequence database at the time of the study. However, following the study period, all isolates were subjected to rpoB gene sequence analysis, which is required to differentiate the species from M. szulgai, and no isolates were identified as M. angelicum.

^cOther members of the M. chelonae complex, M. salmoniphilum, M. franklinii, and M. saopaulense, were not tested.

^dM. farcinogenes is a slowly growing Mycobacterium species.

^eThe other member of the M. smegmatis group, M. goodii, was not tested.

^fThe other closely related member, M. wolinskyi, was not tested.

^gThe other member of the M. mucogenicum group, M. aubagnense, was not tested.

^hMost closely related.

TABLE 2.

Accuracy of identification of nocardia

Reference identificationa	No. (%) of isolates with indicated Vitek MS v3.0 ID result
	Correct to species level	Correct to complex or group level	Incorrect	No ID
N. abscessus (n = 16)	15 (94)			1 (6)
N. arthritidisN (n = 6)			1 (17)	5 (83)
N. asiaticaN (n = 5)			4 (80)	1 (20)
N. beijingensisN (n = 1)			1 (100)
N. brevicatena/paucivorans complex
N. paucivorans (n = 18)	18 (100)
N. nova complex
N. nova (n = 33)		33 (100)
N. veterana (n = 23)	23 (100)
N. elegansN (n = 2)		2 (100)
N. transvalensis complex
N. transvalensis (n = 22)	22 (100)
N. wallacei (n = 13)	13 (100)
N. blacklockiaeN (n = 6)		6 (100)
N. asteroides (n = 19)	15 (79)		3 (16)	1 (5)
N. cyriacigeorgica (n = 33)	33 (100)
N. farcinica (n = 32)	32 (100)
N. rhamnosiphilaN (n = 2)				2 (100)
N. testaceaN (n = 1)				1 (100)
N. brasiliensis (n = 37)	35 (95)			2 (5)
N. vulnerisN (n = 1)			1 (100)
N. pseudobrasiliensis (n = 13)	12 (92)			1 (8)
N. mexicanaN (n = 1)				1 (100)
N. otitidiscaviarum (n = 18)	18 (100)
Nocardia species (n = 10)		3 (30)		7 (70)
Total (n = 312)	236 (76)	44 (14)	10 (3)	22 (7)

^aA superscript “N” indicates species not included in the v3.0 database.

Performance for the clinical isolates of mycobacteria.

As shown in Table 1, 45 (100%) of the 45 isolates of M. tuberculosis complex were correctly identified. Ninety-two percent of the slowly growing NTM and 98% of the rapidly growing mycobacteria (RGM) were correctly identified to the species, complex, or group level. Among the RGM, the Vitek system was able to accurately identify species within the M. chelonae/abscessus group as M. chelonae (28 of 29 correctly identified to the species level, 1 “no identification”), M. immunogenum (19 of 19 correctly identified to the species level), or M. abscessus complex (40 of 40 correctly identified). The Vitek system was unable to differentiate the three subspecies of M. abscessus (i.e., subspecies abscessus, massiliense, and bolletii). Although the Vitek system cannot differentiate between species within the M. fortuitum group, 48 (98%) of 49 isolates tested within this group were identified correctly. Within the M. mucogenicum group, 18 of 20 isolates were correctly identified as M. mucogenicum. However, M. phocaicum, which is not represented in the database, had 20 of the 21 isolates misidentified as M. mucogenicum. The remaining M. phocaicum isolate was not identified. Additionally, there was one isolate identified by the reference method as being most closely related to M. mucogenicum/phocaicum that was identified as M. mucogenicum by the Vitek system.

Among the slowly growing NTM, eight species were accurately identified to the species level for every isolate tested; this included 28 of 28 M. simiae, 31 of 31 M. kansasii, 25 of 25 M. marinum, 15 of 15 M. scrofulaceum, and 24 of 24 M. xenopi isolates. The Vitek system was relatively accurate in identifying the other slowly growing NTM species, including 28 (93%) of 30 M. gordonae, 17 (94%) of 18 M. haemophilum, and 23 (96%) of 24 M. szulgai isolates. The Vitek system was also able to successfully identify the more commonly encountered members of the M. avium complex (MAC), namely, M. avium (68 of 68; 100% accurate) and M. intracellulare (42 of 43; 98% accurate). There was one isolate of M. intracellulare that was misidentified as M. avium (Table 3). Other MAC members were misidentified at the species level (Table 3) but correctly identified at the complex level, generally as a result of not being included in the Vitek database. These included 16 M. chimera isolates, 1 of which was identified as M. avium and 15 of which were identified as M. intracellulare. Similarly, 14 “MAC-X” isolates (49) were identified as M. intracellulare, while an additional six isolates could not be identified. A result of “no identification” was obtained for all three isolates of M. shigaense, the single member of the M. simiae complex not represented in the database. There were three additional slowly growing NTM species that were not represented in the Vitek database, including M. paraense (one isolate was not identified), M. parascrofulaceum (one isolate was not identified and one isolate was misidentified as M. scrofulaceum), and M. paraffinicum (of nine isolates, seven were not identified and two were misidentified as M. scrofulaceum). Considering all of the mycobacterial isolates tested, the Vitek system successfully identified 94% to the species, complex, or group level.

TABLE 3.

Identifications correct to the complex or group level

Reference identificationa	Vitek MS v3.0 identification(s) (no. of isolates)
Mycobacteria
M. tuberculosis complex	M. tuberculosis complex (45)
M. intracellulare	M. avium (1)
M. chimeraN	M. avium (1), M. intracellulare (15)
MAC-XN,b	M. intracellulare (14)
M. gordonae/paragordonaeN	M. gordonae (1)
M. abscessus complex	M. abscessus (40)
M. fortuitum group	M. fortuitum group (48)
M. phocaicumN	M. mucogenicum (20)
M. mucogenicum/phocaicumc	M. mucogenicum (1)
Nocardia
N. nova	N. nova/africana (33)
N. elegansN	N. nova/africana (2)
N. blacklockiaeN	N. wallacei (6)

^aA superscript “N” indicates species not included in the v3.0 database.

^bMAC-X includes M. arosiense, M. bouchedurhonense, M. colombiense, M. marseillense, M. timonense, M. vulneris, and M. yongonense.

^cMost closely related.

Performance for the clinical isolates of Nocardia.

A total of 312 isolates representing 21 different Nocardia species were tested (Table 2). Of these, 236 (76%) were correctly identified to the species level and another 44 (14%) were correctly identified to the complex level. Twenty-two (7%) isolates could not be identified despite repeat testing. Seven common species (N. cyriacigeorgica, N. farcinica, N. otitidiscaviarum, N. paucivorans, N. transvalensis, N. veterana, and N. wallacei) were correctly identified to the species level for every isolate tested (33/33, 32/32, 18/18, 18/18, 22/22, 23/23, and 13/13, respectively). Fifteen (79%) of 19 N. asteroides isolates were identified to the species level by the Vitek system. Three N. asteroides isolates were misidentified as N. neocaledoniensis (Table 4), and one was not identified. While all 23 N. veterana isolates were correctly identified to the species level, the Vitek system did not perform as well on other members of the N. nova complex. Each of the 33 N. nova isolates was given a result consisting of both N. nova and N. africana identifications (Table 3). This was also the case for the two isolates of N. elegans, which, notably, is not represented in the Vitek database (Table 3). Similarly, all 22 isolates of N. transvalensis and all 13 isolates of N. wallacei were correctly identified to the species level, but the other member of the N. transvalensis complex, N. blacklockiae, which is not in the database, was consistently (6 of 6, 100%) misidentified as N. wallacei. There were six additional species that were not represented in the Vitek database. This included two isolates of N. rhamnosiphila and one isolate of N. testacea, which were not identified; one isolate of N. asiatica and one isolate of N. beijingensis, which were misidentified as N. abscessus; two isolates of N. asiatica and one isolate of N. arthritidis, which were misidentified as N. beijingensis; one isolate of N. asiatica, which was misidentified as N. abscessus/N. beijingensis, and one isolate of N. vulneris, which was misidentified as N. brasiliensis. Considering all Nocardia isolates tested, 280 (90%) were correctly identified to the species or complex level.

TABLE 4.

Incorrect identifications

Reference identificationa	Vitek MS v3.0 identification(s) (no. of isolates)
Mycobacteria
M. gordonae/paragordonaeN	M. scrofulaceum (1)
M. parascrofulaceumN	M. scrofulaceum (1)
M. paraffinicumN	M. scrofulaceum (2)
Nocardia
N. asiaticaN	N. abscessus (1), N. beijingensis (2), N. abscessus/beijingensis (1)
N. beijingensisN	N. abscessus (1)
N. asteroides	N. neocaldonensis (3)
N. arthritidisN	N. beijingensis (1)
N. vulnerisN	N. brasiliensis (1)

^aA superscript “N” indicates species not included in the v3.0 database.

Repeat testing.

Repeat testing was performed when the Vitek gave an initial result of “no identification” or when the spectra were of poor quality and a result was not provided. In either case, the original extract was respotted as a single replicate. If no identification was made again, a new sample extract was prepared from the same subculture and tested as a single replicate. If the identification could still not be obtained, a result of “no identification” was used as the final result. As shown in Table 5, repeat testing was performed on 103 isolates, resulting in an overall repeat rate of 11% (103 of 963). The repeat rate was slightly lower for the mycobacteria (68 of 651, 10%) than for the Nocardia (35 of 312, 11%). Species not represented in the v3.0 database typically had the highest repeat rates. For those species in the database, the repeat testing rate was highest for M. lentiflavum (35%). and N. abscessus (44%). One isolate was misidentified during repeat testing; an N. asteroides was misidentified as N. neocaledoniensis after respotting. There were three species, M. haemophilum, M. simiae, and M. xenopi, for which repeat testing was performed solely due to poor-quality spectra as indicated by the software. None of the Nocardia isolates required repeat testing as a result of poor-quality spectra.

TABLE 5.

Results of repeat testing

Reference identificationa	No. of isolates requiring repeat testing (repeat rate, %)	No. of isolates with indicated Vitek MS identification result
		Respot correct	Reextract correct	Reextract incorrect	Final
Mycobacteria (total tested = 813)
M. abscessus complex (n = 40)	1 (3)	1
M. avium (n = 68)	3 (4)	1	2
M. chelonae (n = 29)	1 (3)				1
MAC-XN (n = 20)	6 (30)				6
M. fortuitum group (n = 49)	1 (2)				1
M. gordonae (n = 33)	6 (17)	4			2
M. gordonae/paragordonaeN (n = 3)	1 (100)				1
M. haemophilum (n = 18)	3 (9)	1	1		1
M. kansasii (n = 31)	4 (13)	4
M. lentiflavum (n = 37)	13 (30)	4	5		4
M. malmoense (n = 4)	4 (12)	2
M. mucogenicum (n = 20)	4 (13)	1	1		2
M. paraenseN (n = 1)	1 (100)				1
M. parascrofulaceumN (n = 2)	1 (50)				1
M. paraffinicumN (n = 9)	7 (78)				7
M. phocaicumN (n = 21)	1 (5)				1
M. scrofulaceum (n = 15)	3 (10)	1
M. shigaenseN (n = 3)	3 (100)				3
M. simiae (n = 28)	1 (3)		1
M. smegmatis (n = 21)	6 (20)	1	3
M. szulgai (n = 24)	2 (6)	1			1
M. xenopi (n = 24)	1 (3)		1
Mycobacterium speciesN (n = 1)	1 (100)				1
Total	68 (10)	21	14		33
Nocardia (total tested = 417)
N. abscessus (n = 16)	7 (23)		6		1
N. arthritidisN (n = 6)	5				5
N. asiaticaN (n = 5)	1				1
N. asteroides (n = 19)	2 (6)			1	1
N. brasiliensis (n = 37)	2 (5)				2
N. mexicanaN (n = 1)	1				1
N. otitidiscaviarum (n = 18)	1 (3)	1
N. pseudobrasiliensis (n = 13)	4 (13)	3			1
N. rhamnosiphilaN (n = 2)	2				2
N. testaceaN (n = 1)	1				1
N. vulnerisN (n = 1)	1			1
N. wallacei (n = 13)	1 (3)	1
Nocardia speciesN (n = 10)	7				7
Total	35 (11)	5	6	1	22
Total (n = 963)	103 (11)	26	20	2	55

^aA superscript “N” indicates species not included in the v3.0 database. Only those species that had at least one isolate that required repeat testing are listed.

Challenge set.

In addition to the isolates obtained during the course of clinical work, a panel of 50 well-characterized isolates representing 18 Mycobacterium and 12 Nocardia species were run at each of three sites. As shown in Table 6, 149 150 (99%) of the replicates were correctly identified. There were no misidentifications, but three N. nova isolates were identified as N. nova/N. africana and thus correctly identified to the complex level rather than the species level. One M. haemophilum isolate was not identified at a single site.

TABLE 6.

Accuracy of Vitek MS v3.0 for the challenge set

Organism (no. of isolates)	No. (%) of isolates with indicated Vitek MS identification result
	Correct to species level	Correct to complex or group level	Incorrect	No ID
M. abscessus complex (n = 6)		6 (100)
M. avium (n = 9)	9 (100)
M. chelonae (n = 3)	3 (100)
M. fortuitum group (n = 15)		15 (100)
M. gordonae (n = 3)	3 (100)
M. haemophilum (n = 6)	5 (83)			1 (17)
M. immunogenum (n = 3)	3 (100)
M. intracellulare (n = 9)	9 (100)
M. kansasii (n = 9)	9 (100)
M. malmoense (n = 3)	3 (100)
M. marinum (n = 3)	3 (100)
M. mucogenicum (n = 3)	3 (100)
M. scrofulaceum (n = 3)	3 (100)
M. simiae (n = 3)	3 (100)
M. smegmatis (n = 3)	3 (100)
M. szulgai (n = 3)	3 (100)
M. tuberculosis complex (n = 15)		15 (100)
M. xenopi (n = 6)	6 (100)
N. abscessus (n = 6)	6 (100)
N. asteroides (n = 6)	6 (100)
N. brasiliensis (n = 3)	3 (100)
N. cyriacigeorgica (n = 3)	3 (100)
N. farcinica (n = 3)	3 (100)
N. nova (n = 3)		3 (100)
N. otitidiscaviarum (n = 3)	3 (100)
N. paucivorans (n = 3)	3 (100)
N. pseudobrasiliensis (n = 3)	3 (100)
N. transvalensis (n = 3)	3 (100)
N. veterana (n = 6)	6 (100)
N. wallacei (n = 3)	3 (100)
Total no. of isolates in challenge set (n = 150)	110 (73)	39 (26)		1 (1)

Reproducibility.

Reproducibility was assessed by testing a panel of five organisms (M. abscessus, M chelonae, M. smegmatis, N. wallacei, and N. otitidiscaviarum) in duplicate on two runs a day for 5 days at each of three sites for a total of 300 replicates. This reproducibility testing produced a single correct identification at the species level for 297 (99%) of the replicates tested. In two instances, M. smegmatis was not identified; both cases occurred at a single site. There was also a single instance where N. wallacei was not identified at a single site. There were no misidentifications of any organism at any site. There also were no differences in the accuracy of organism identification by site, operator, reagent kit, or day.

DISCUSSION

The use of MALDI-TOF mass spectrometry (MS) to identify commonly encountered isolates of Gram-positive and Gram-negative bacteria and yeasts has revolutionized workflow and improved turnaround time in clinical microbiology laboratories around the world (36, 38). MALDI-TOF MS has previously been applied to Mycobacterium and Nocardia species with mixed results, raising questions about the utility of this method for these organisms (17–21, 23–28, 32, 34–40). Previous studies have used a range of culture conditions, different biomass volumes, and/or different inactivation and extraction procedures (16, 18, 20, 24, 29–31, 33). Several studies have suggested that manufacturer-derived databases require the addition of custom libraries to enable species-level identification rates of at least 90% (18, 20, 25, 26, 39, 40). Given that custom libraries require extensive development and validation, their use may not be a viable option for many clinical laboratories. These factors have collectively led to an absence of more widely spread use of MALDI-TOF mass spectrometry for the identification of these bacteria.

This report provides evidence that results obtained using the Vitek system are accurate and reproducible for the identification of Mycobacterium and Nocardia species. The v3.0 system is an improvement over the previous system in that it uses a simple and optimized extraction procedure specifically designed for the mycobacteria and Nocardia and relies on a well-developed database. The optimized extraction procedure has also been shown to be effective at inactivating infectious organisms, such as members of the M. tuberculosis complex (22). The ability of the Vitek system to identify M. tuberculosis complex with high (100%) accuracy is especially important for patient care and infection control. In this study, there were very few (1%) misidentifications. There was also a relatively low frequency of “no identification” calls (6%) which was similar to the rate seen with other bacterial species tested using this system (range, 1% to 8%) (43, 50–53). In the work presented here, isolates that failed to obtain an identification on the first attempt were respotted and then reextracted before the final result of “no identification” was accepted. Using this strategy, almost half (47%) of the isolates with an initial result of “no identification” were able to be correctly identified. Given that the repeat testing rate was low (11%), this strategy may be more efficient than duplicate spotting during the first attempt at identification.

The strengths of this study were the large number of isolates tested, representing a variety of different taxa; the use of a standardized inactivation and extraction procedure; the availability of a well-curated database; and use of a formalized repeat testing algorithm. One limitation of this study was that some clinically important organisms such as members of the M. terrae complex, as well as rarer species within the M. chelonae complex, M. smegmatis, and M. mucogenicum groups, were not thoroughly tested. Another limitation was that all testing was performed using isolates from growth on solid media. Additional studies evaluating the accuracy of the Vitek system for organisms grown in liquid media are under way.

Although the Vitek MS v3.0 system can differentiate some closely related species, it is unable to do so for others. Members of the MAC, M. tuberculosis, and M. abscessus complex; the M. fortuitum and M. mucogenicum groups; and the N. nova complex frequently could not be differentiated. Similarly, although the system correctly identified all M. avium isolates, other members of the M. avium complex could not be reliably differentiated. This can be problematic for species not included in the database that are closely related to those included in the database, such as M. chimaera, as they may be misidentified as a similar species rather than not identified. Furthermore, although 41 of 41 M. abscessus isolates were correctly identified to the species level, the Vitek MS v3.0 system cannot differentiate between the closely related subspecies within this group, including M. abscessus subspecies abscessus, bolletii, and massiliense. These subspecies are of particular concern since they can vary in their levels of resistance to macrolides due to the presence or absence of a functional erm(41) gene. Reporting identifications to the complex or group level may be warranted in these cases. Similarly to the mycobacteria, the use of molecular methods has revised the taxonomy of Nocardia during the last 20 years, with many new species and complexes being defined (1). Selection of optimal antimicrobial therapy for infections caused by Nocardia is dependent on correct identification to the species level. However, identification to the complex level, such as was the case for species within the N. nova and N. transvalensis complexes, may be sufficient given their similar drug susceptibility profiles (1).

In summary, using standardized methods for growth, sample preparation, and analysis, the Vitek MS v3.0 system was 93% accurate for the identification of Mycobacterium and Nocardia isolates to the species, group, or complex level. Using this system, it may now be feasible for many clinical laboratories to accurately identify the commonly encountered mycobacteria and Nocardia.