Introduction

Development of the “soft” ionization technique called matrix-assisted laser desorption/ionization (MALDI) in 1985 allowed the mass spectrometric detection of macromolecules without fragmentation [1–3]. A new approach to matrix composition with time-of-flight (TOF) mass spectrometry in 1988 enabled large protein molecules to be ionized [4]. Advances in microbial genomic research led to the recognition that many mass spectral peaks represent ribosomal proteins, and this proteomic technology could be used to identify clinically important bacteria such as Enterobacteriaceae [5, 6]. Reference mass spectra libraries and software developed for data analysis have been incorporated into commercial MALDI-TOF mass spectrometry identification systems for use in clinical laboratories.

This multicenter study evaluated the accuracy of the VITEK MS system (bioMérieux, Marcy l’Etoile, France) for the mass spectrometric identification of Enterobacteriaceae typically encountered in the clinical laboratory. Because organisms in the Enterobacteriaceae family are biochemically active, automated and manual phenotypic identification systems perform well, but require up to 48 h for the results to be available. Mass spectrometry can provide rapid identifications that are available within minutes [3, 5].

The Enterobacteriaceae family is a heterogeneous group of bacteria ranging from species that are part of normal intestinal flora to organisms that are always considered pathogens. Diseases associated with Enterobacteriaceae include urinary tract, intestinal, respiratory, wound, and bloodstream infections [7]. Antimicrobial-resistant Enterobacteriaceae that produce carbapenemases or extended-spectrum β-lactamases (ESBLs) (e.g., Klebsiella pneumoniae, Klebsiella oxytoca, Enterobacter spp., Escherichia coli) have become a public health concern [7, 8].

This is the first report of VITEK MS system performance for Enterobacteriaceae identification using a new database (v2.0) and MYLA software developed for in vitro diagnostic (IVD) use. The VITEK MS software has been improved by the selective weighting of mass spectral peaks, a feature that is especially helpful for distinguishing between species. The laser scanning function of the IVD system has been optimized to allow more variability in the amount of organism applied to the target slide.

Materials and methods

Bacterial isolates

Study sites were asked to test a minimum of ten Enterobacteriaceae isolates each for common species and six isolates representing less common Enterobacteriaceae species from unique patients. If an insufficient number of isolates were available at a study site, bioMérieux provided stock isolates. The five participating study sites were: Cleveland Clinic (Cleveland, OH), Massachusetts General Hospital (Boston, MA), Barnes Jewish Hospital (St. Louis, MO), North Shore-LIJ Health System Laboratories (Lake Success, NY), and UCLA (Los Angeles, CA). Each study site obtained institutional review board approval prior to the initiation of the clinical trial.

MALDI-TOF mass spectrometry

Clinical isolates were analyzed on the VITEK MS IVD system (database v2.0), in accordance with manufacturer’s instructions. The E. coli ATCC 8739 strain was used for every acquisition group on the target slide to calibrate the mass spectrometer. For a negative control, matrix solution (α-cyano-4-hydroxycinnamic acid; bioMérieux) was tested alone. In addition, one of four quality control organisms (Enterobacter aerogenes ATCC 13048, K. oxytoca ATCC 13182, Pseudomonas aeruginosa ATCC 10145, Staphylococcus aureus ATCC 29213) was tested with each new lot of target slides and matrix solution, as well as on each day of clinical isolate testing.

The protocol allowed isolates to be tested as long as 72 h after subculture. Most isolates were tested using ≤48 h growth from trypticase soy agar with 5 % sheep blood (Remel, Lenexa, KS). Only nine cultures were >48 h growth when tested. Thirty-eight isolates were tested from MacConkey II MUG agar (BBL, Sparks, MD). Frozen isolates were subcultured twice before mass spectrometric analysis. A thin layer of colony growth was applied directly to the target slide using a 1-μl loop. Matrix solution (1 μl) was added and allowed to dry before mass spectrometric analysis. On the basis of the confidence level, the VITEK MS system provided a species, genus only, or no identification for each isolate. Repeat VITEK MS testing was performed if there was quality control failure, calibration failure, poor or absent mass spectra, technical error, or a mixed culture.

Reference identification

Growth from the plate used for VITEK MS testing was inoculated to slants shipped for reference testing at MIDI Labs (Newark, DE) and bioMérieux. At MIDI Labs, 16S rRNA gene sequencing was performed using the MicroSEQ 500 16S rDNA Bacterial Identification Kit (Applied Biosystems, Foster City, CA) with Sherlock DNA software (MIDI) analysis. The sequencing data were also analyzed by researchers at bioMérieux using the GenBank database [9] and BIBI (Bioinformatics Bacterial Identification) software [10]. Final 16S rRNA gene sequencing identifications were assigned according to Clinical and Laboratory Standards Institute (CLSI) interpretive criteria [11]. Supplemental phenotypic testing was performed at bioMérieux with VITEK 2 GN cards (bioMérieux), API 20E strips (bioMérieux), and/or classical biochemical tube or spot tests for organisms unidentified by 16S rRNA gene sequencing.

Data analysis

Each VITEK MS result (species or genus level only) was compared to the final reference identification and classified as correct if concordant or a misidentification if discordant. Isolates with a reference identification that is not claimed by VITEK MS were excluded from the study.

Results

Enterobacteriaceae isolates representing 17 genera and 40 species were included in the study (Table 1). The majority of the 965 study isolates (73.1 %) were recovered from patient cultures performed at one of the five study sites. The remaining 260 isolates (26.9 %) were unique isolates provided by bioMérieux representing rare (17.9 %) or uncommon (9 %) strains.

Table 1 Comparison of the VITEK MS results to reference method identifications for 965 Enterobacteriaceae isolates
Full size table

The accuracy of VITEK MS identifications in comparison to the reference method is shown in Table 1. The VITEK MS result agreed with the reference method for 96.7 % of the 965 isolates tested, with 83.8 % correct to the species level and 12.8 % limited to a genus-level identification. A small percentage of isolates (1.7 %) were not identified by VITEK MS. Details of the 15 VITEK MS results (1.5 %) classified as misidentifications are shown in Table 2.

Table 2 Results for the 15 Enterobacteriaceae isolates misidentified by VITEK MS
Full size table

Optimal performance (100 % of isolates assigned a VITEK MS species-level identification concordant with the reference method) occurred for Citrobacter koseri (n = 31), E. aerogenes (n = 52), Enterobacter gergoviae (n = 10), Ewingella americana (n = 6), K. oxytoca (n = 49), K. pneumoniae (n = 58), Morganella morganii (n = 52), Providencia stuartii (n = 31), Serratia marcescens (n = 57), Serratia odorifera (n = 30), Yersinia enterocolitica (n = 14), Yersinia intermedia (n = 9), and Yersinia pseudotuberculosis (n = 8). There was also 100 % species-level agreement for the 65 E. coli tested, but this result includes a limitation that MALDI-TOF mass spectrometry cannot differentiate between E. coli and Shigella spp. and, for this reason, Shigella spp. were not included in the study. Five additional organisms had species-level identifications concordant with the reference method for at least 90 % of isolates: Citrobacter amalonaticus (90 %), Leclercia adecarboxylata (90 %), Proteus mirabilis (98.3 %), Providencia rettgeri (97 %), Salmonella enterica (94.3 %), and Serratia liquefaciens (95.7 %).

Seven organisms were more likely to have a genus- rather than species-level VITEK MS identification: Citrobacter braakii (33.3 % species, 44.4 % genus), Citrobacter youngae (38.5 % species, 61.5 % genus), Enterobacter asburiae (0 % species, 83.3 % genus), E. cloacae (0 % species, 96.3 % genus), Proteus penneri (0 % species, 94.7 % genus), and Proteus vulgaris (0 % species, 100 % genus). The species options included for genus-level VITEK MS identifications are shown in Table 3.

Table 3 Results for the 124 Enterobacteriaceae isolates with VITEK MS identification limited to the genus level
Full size table

There were only two organisms, Citrobacter freundii and Pantoea agglomerans, with multiple VITEK MS misidentifications. Four C. freundii isolates had an incorrect VITEK MS species identification of C. youngae (n = 2) or Citrobacter werkmanii (n = 2). Three P. agglomerans isolates were misidentified by VITEK MS as Enterobacter cancerogenus (n = 2) or Enterobacter spp. (Table 2).

Supplemental phenotypic testing to determine the final reference identification was required for 167 of the 964 isolates. In addition to 16S rRNA gene sequencing, biochemical testing using VITEK 2 GN cards, API 20E strips, and/or classical tube or spot tests was performed on eight C. amalonaticus, eight C. braakii, 19 C. freundii, three C. koseri, four C. youngae, 29 E. aerogenes, three E. asburiae, one E. cancerogenus, 18 E. cloacae, six Escherichia coli, one E. hermannii, seven Hafnia alvei, 20 K. oxytoca, two K. pneumoniae, one Morganella morganii, four P. agglomerans, one Proteus mirabilis, 17 P. vulgaris, three Providencia rettgeri, one Providencia stuartii, one Raoultella ornithinolytica, one Raoultella planticola, three Salmonella enterica, three Serratia liquefaciens, two S. marcescens, and one Yersinia enterocolitica isolate.

Discussion

This multicenter study demonstrated that identifications provided by the VITEK MS IVD system for Enterobacteriaceae are highly accurate in comparison to a molecular reference method. The Enterobacteriaceae family includes genera and species that can be difficult to identify using 16S rRNA gene sequencing, so supplemental phenotypic testing was required for many isolates [11–13]. Some limitations of 16S rRNA sequencing identification have also been observed with mass spectrometry methods. For example, E. coli and Shigella spp. are indistinguishable by 16S rRNA gene sequencing and the VITEK MS system. It has been suggested that Shigella and E. coli should be reclassified as the same species [14, 15]. Mass spectrometric identifications of E. coli can be accepted for isolates showing lactose fermentation and indole production or motility, but serotyping, lysine decarboxylation, and other phenotypic tests are required in order to differentiate Shigella from inactive (lactose-nonfermenting) E. coli [16]. The recovery of Shigella from an extraintestinal site is uncommon [16].

Genera such as Citrobacter, Enterobacter, Klebsiella, and Pantoea belong to multiple genogroups and their taxonomic classification is based on biochemical testing [11]. Despite this genetic heterogeneity, the overall accuracy of VITEK MS identifications for these genera was excellent with only a small number of misidentifications. For multiple Citrobacter and Enterobacter species, VITEK MS identifications limited to the genus level were common.

The most frequent VITEK MS genus misidentification was for P. agglomerans: 13.6 % of isolates were reported as species in the closely related genus Enterobacter. Yellow pigmented colonies with a mass spectrometric identification of Enterobacter spp. should be assessed for the biochemical characteristics of P. agglomerans (negative reactions for ornithine decarboxylase and arginine dihydrolase) and Cronobacter sakazakii [7]. Although most P. agglomerans are usually less antimicrobial resistant than Enterobacter spp., a recent report of an ESBL-producing strain highlights the need for susceptibility testing of clinically significant Enterobacteriaceae [17].

The majority of VITEK MS species misidentifications were within the Citrobacter genus. Four C. freundii isolates were reported as C. youngae or C. werkmanii, and one C. braakii was misidentified as C. freundii. Most C. youngae and C. braakii were genus- rather than species-level identifications, with the result split between multiple species. No isolates with a reference identification of C. werkmanii were included in this study. Use of the term “Citrobacter freundii complex” for C. freundii, C. youngae, C. braakii, C. werkmanii, and C. sedlakii [18] may be the best approach to reporting mass spectrometry results for these species. The C. amalonaticus and C. koseri isolates studied were reliably identified to the species level by VITEK MS and tend to be more susceptible to cephalosporins than C. freundii.

Six species are included in the Enterobacter cloacae complex (E. asburiae, E. cloacae, E. hormaechei, E. kobei, E. ludwigii, E. nimipressuralis) [19]. The only E. cloacae complex species included in this study were E. asburiae (n = 10) and E. cloacae (n = 26). Since the VITEK MS system provided genus-level identifications for all 36 isolates as a split between these two species, laboratories may choose to report “E. cloacae complex” rather than “Enterobacter spp.”

The VITEK MS results for C. koseri, E. aerogenes, E. gergoviae, E. coli, E. americana, K. oxytoca, K. pneumoniae, M. morganii, P. mirabilis, P. rettgeri, P. stuartii, S. marcescens, S. odorifera, Y. enterocolitica, Y. intermedia, and Y. pseudotuberculosis were excellent, with 97–100 % of isolates correctly identified to the species level. The inability of VITEK MS to discriminate between P. vulgaris and P. penneri can be resolved with supplemental indole testing.

There are five options for Salmonella enterica reporting by the VITEK MS v2.0 database: Salmonella serotype Typhi, Salmonella serotype Paratyphi A, Salmonella serotype Gallinarum, Salmonella enterica subsp. arizonae/Salmonella enterica subsp. diarizonae, and Salmonella group (Salmonella enterica subsp. enterica, Salmonella serotype Enteritidis, Salmonella serotype Paratyphi B, Salmonella serotype Paratyphi C, Salmonella serotype Typhimurium). The limited variety of Salmonella isolates included in this study did not allow VITEK MS identification of specific serotypes to be assessed. The utility of mass spectrometry to prescreen for some important Salmonella enterica serovars and reduce the number of isolates requiring traditional serotyping at public health laboratories has been demonstrated in other studies [20, 21].

Because the clinical significance and susceptibility test methods are similar for Enterobacteriaceae, the impact of the VITEK MS misidentifications reported in this study on patient care would be minimal. Most laboratories have routine procedures in place requiring any unusual identification (such as the Obesumbacterium proteus result that occurred for the H. alvei isolate) to be followed up with additional biochemical testing.

Cherkaoui et al. reported a higher percentage of correct mass spectrometric Enterobacteriaceae identifications provided by the Bruker Biotyper (Bruker Daltonics) in comparison to the Shimadzu system (current VITEK MS RUO) [22]. In the current study, the accuracy of K. pneumoniae, E. cloacae, and S. marcescens identifications provided by the VITEK MS IVD system were better than the prior report of Shimadzu system performance and similar to the performance of the Bruker Biotyper [22].

Saffert et al. reported a lower percentage of S. marcescens, K. oxytoca, and P. agglomerans isolates identified to the genus or species level by the Bruker Biotyper (v2.0 software) in comparison to the current VITEK MS study [23]. The Bruker Biotyper had difficulty differentiating Klebsiella from the closely related genus Raoultella [23, 24].

A prospective Bruker Biotyper study of all clinical isolates identified during a four-week period included 15 Enterobacteriaceae species (∼50 % were E. coli) [25]. Bruker (v3.0 software) misidentifications for Shigella, Citrobacter, Enterobacter, and Klebsiella species, similar to the present VITEK MS study, were noted [25].

Martiny et al. compared the VITEK MS RUO, VITEK MS IVD, and Microflex LT/Bruker Biotyper (v3.0 software) systems available in Europe by testing many E. coli and small numbers of other species in the Enterobacteriaceae family [26]. The IVD version of each system performed well, but higher errors for the species-level identification of Serratia species were noted with the Bruker database [26].

Strengths of our study include the large number of Enterobacteriaceae isolates from different geographic regions of the U.S. and the use of a reference molecular method for all isolates. A study limitation is the lack of clinically important species such as Shigella spp. and a variety of Salmonella serotypes (e.g., Typhi, Paratyphi A).

The efficient workflow used in our study (single spotting of isolates, no extraction step) minimizes the hands-on time required for mass spectrometric testing. Although not measured in this study, the faster identification available with mass spectrometry should improve patient care [22, 27–30]. Lower reagent and labor costs for organism identification is another benefit of MALDI-TOF mass spectrometry [22, 27–30].

In conclusion, the VITEK MS IVD system provided accurate genus- or species-level identifications for a large and diverse collection of Enterobacteriaceae clinical isolates. The implementation of MALDI-TOF mass spectrometric identification will allow laboratories to provide results in a more clinically relevant time frame than current commercial biochemical identification systems.