Multicenter Evaluation of the Bruker MALDI Biotyper CA System for the Identification of Clinically Important Bacteria and Yeasts

Materials and Methods

Infections remain a significant cause of morbidity and mortality in modern medicine, despite the availability of broad-spectrum antimicrobial agents. Five of the most important health care–associated infections, studied by Zimlichman and colleagues¹ in 2013, were estimated to cost the US health care system $9.8 billion. Empiric therapy, which routinely includes broad-spectrum, often expensive, antimicrobial agents, is necessary to provide “coverage” for the most likely pathogens for a particular type of infection prior to the identification of the specific etiologic agent of disease.^2,3 Although often necessary early in the course of infection, the continued use of broad-spectrum antimicrobial agents after pathogen identification is discouraged since this practice contributes to antimicrobial resistance, is associated with Clostridium difficile infections, and is costly.^4,5 Therefore, the prompt and accurate identification of microbial pathogens from clinical specimens in conjunction with antimicrobial stewardship practices is important and offers opportunities to use targeted rather than broad-spectrum antimicrobial therapy while preserving broad-spectrum agents and, in many instances, decreasing health care costs.⁶

Foremost, prompt and accurate identification of microbial pathogens ensures that the empiric coverage selected is appropriate. For example, broad-spectrum antibacterial coverage would be ineffective if it was discovered that a bloodstream infection was secondary to Candida albicans. Rapid identification of the etiologic agent of infection may also disclose the possibility of antimicrobial resistance. For example, the choice of vancomycin for broad-spectrum gram-positive coverage should be questioned if the infecting isolate was discovered to be Enterococcus faecium, a possible vancomycin-resistant organism. In this scenario, the rapid identification could be used to selectively test for the vanA and vanB genes by molecular methods.⁷ Finally, once the identification is known, the tailoring of therapy may begin, such as the selection of antipseudomonal penicillin or cephalosporin, in contrast to other broad-spectrum gram-negative coverage, once the isolate is identified as Pseudomonas aeruginosa.⁸

The traditional approach to bacterial and yeast isolate identification, with the exception of spot tests (eg, spot indole) and rapid testing (eg, germ tube), requires biochemical testing, which requires an additional period of microorganism growth (eg, growth in the presence of hippurate to assess hippurate hydrolysis). Rapid molecular tests decrease the time to a definitive identification, but these tests are often expensive and sometimes require several processing steps.⁹ The recent development of bacterial and yeast identification by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry significantly decreases the time of definitive identification.¹⁰ The MALDI-TOF process has fewer steps than most molecular assays, has among the fastest time-to-result of tests in clinical microbiology, and offers an extremely low cost for identification because very few reagents are used.¹¹ MALDI-TOF mass spectrometry, therefore, represents a significant advancement in clinical microbiology.

The clinical utility and accuracy of MALDI-TOF mass spectrometry has been evaluated by numerous clinical microbiologists on a wide variety of medically important microorganisms.^12‐18 This current study represents the multicenter US Food and Drug Administration (FDA) 510(k) submission trial of the MALDI Biotyper CA System (Bruker Daltonics, Billerica, MA), which includes a large library of clinically important aerobic and anaerobic bacteria and yeasts. The MALDI biotyper and software library evaluated in this study have subsequently been approved for use by the FDA.

Results

Participating Sites

Five clinical sites and Bruker Daltonics (Bruker Daltonik, Bremen, Germany) participated in this clinical trial for a 510(k) submission to the FDA. The five clinical trial sites were Cleveland Clinic, Cleveland, Ohio; Dynacare Laboratories (now Wisconsin Diagnostic Laboratories), Milwaukee, Wisconsin; Kaiser Permanente Southern California Permanente Medical Group Regional Reference Laboratories, North Hollywood; Laboratory Alliance of Central New York, Syracuse; and TriCore Reference Laboratories, Albuquerque, New Mexico. All sites received approval from their respective institutional review boards for the performance of these studies.

Isolate Collection and Assessment

In total, 4,399 isolates of medically important bacteria and yeasts were included in this study Table 1. These included 3,584 bacterial isolates and 815 yeast isolates. The bacterial isolates consisted of 2,262 aerobic gram-positive bacteria, 792 aerobic gram-negative bacteria, and 530 gram-positive and gram-negative anaerobic bacteria. The yeasts consisted of predominantly of Candida species (601 isolates representing 22 species), as well as 75 isolates representing three species or variants of Cryptococcus and 139 isolates of six other yeast or yeast-like fungal genera. All of the isolates were submitted for ribosomal DNA (DNA) sequencing using criteria defined in the Clinical Laboratory Standards Institute’s (CLSI) document MM18-A (CLSI, Wayne, PA), in conjunction with the GenBank and/or EzTaxon databases. The 16S rDNA gene was the primary target for the identification of bacteria, whereas the internal transcribed spacer gene was the primary target for the identification of fungi. If definitive identification could not be achieved with rDNA sequencing, then biochemical testing and/or targeted alternate gene sequencing was used to define the species of the organism. The alternate genetic targets were atpA for Actinomyces odontolyticus; atpD for Cronobacter sakazakii; glyA for Campylobacter species; recA for Streptococcus mitis, Streptococcus oralis, and Streptococcus pneumoniae; rpoB for Haemophilus influenzae; sodA for Staphylococcus species; and tufA for Enterococcus species and Streptococcus salivarius.

Fresh isolates were subcultured to ensure purity and to obtain isolated colonies. Frozen isolates were subcultured twice to ensure purity, provide an opportunity for the isolate to recover, and obtain isolated colonies. All bacterial and yeast isolates were incubated a minimum of 18 hours or until sufficient growth for testing was obtained.

Isolate Preparation and Mass Spectrometry

Discussion

Replicate Testing

A Gram stain was prepared for all isolates prior to testing. A single colony from an appropriately incubated plate was inoculated onto two selected positions on a US IVD 48 Spot Target (target) (Bruker Daltonik). Position 1 was overlaid with 1 μL US IVD HCCA portioned matrix solution (Bruker Daltonik); this was the direct transfer method (DT), as previously described.¹⁹ Position 2 was similarly inoculated but then overlaid with 1 μL 70% formic acid and, after drying, overlaid with 1 μL matrix solution; this was the extended direct transfer method (eDT). The prepared target was processed on the MALDI Biotyper CA System per the manufacturer’s instructions. If either the DT- or eDT-processed samples returned a matching log (score) of less than 2.0, then the isolate was processed with a full extraction method.

The extraction method began with suspending colonies obtained with a 1-μL loop in 300 μL high-performance liquid chromatography–grade water. Using a pipette, 900 μL absolute proof ethanol was added, and the preparation was mixed with pipetting action. A pellet was prepared by centrifugation at 13,000 to 15,000 rpm for 2 minutes, and the supernatant was removed and discarded. A centrifugation step was repeated, and any remaining supernatant was removed using a fine-tip disposable pipet. The pellet was resuspended in 25 μL 70% formic acid and disrupted by tituration. Then, 25 μL 100% acetonitrile was added, and the suspension was mixed in a similar fashion. This mixture was centrifuged, as described above. Then, 1 μL supernatant was pipetted onto the target well, air dried, and then followed by the addition of 1 μL matrix solution. This was air dried and processed on the MALDI Biotyper CA System per the manufacturer’s instructions.

Comparisons were made between the processing methods (eg, DT, eDT, and extraction), alone and in select combinations (ie, [DT + eDT] and [DT + extraction]), and the categories of MALDI scores generated (ie, high level of confidence [≥2.0], low level of confidence [1.70-1.99], and no identification [<1.70]). Categorical comparisons of the processing methods were made with respect to the frequency of categorizing isolates with a high level of confidence using standard statistical methods (EpiCalc 2000; BrixtonHealth [http://www.brixtonhealth.com/epicalc.html]).

For 110 of 140 bacterial species and 28 of 31 yeast species, additional replicate testing was performed to supplement the test data. This replicate testing was carried out if a lower number of isolates could be collected during the studies. Several strains of a species were selected, cultivated, and measured at different test sites for replicate testing. Depending on the number of strains for a species, up to 10 different strains were selected and up to 10 measurements per strain were made at different test sites. Overall, 3,123 additional replicate assessments for bacteria and 679 for yeast were made, for a combination of 3,802 replicate assessments.

Overall, the MALDI Biotyper CA System performed well in the assessment of bacterial and yeast isolates Table 2. In total, 98.4% (4,329/4,399) of the bacterial and yeast isolates were correctly identified to the genus and species level. A significant percentage (ie, 95.7%; 4,209/4,399) of all isolates were identified to the genus and species level with a high degree of confidence, matching a log score of  2.00 or more, whereas only a small percentage (ie, 2.7%; 120/4,399) was identified to the species level with a low degree of confidence, matching a log score between 1.70 and 1.99.

The correct genus but incorrect species or complex occurred in 0.45% (20/4,399) of the isolates; 0.34% (15/4,399) of these were categorized with a high degree of confidence, whereas approximately 0.16% (7/4,399) were categorized with a low degree of confidence (Table 2) Table 3. Approximately 0.18% (8/4,399) of the isolates were assigned the incorrect genus designation; most of these (6/4,399) were erroneously categorized with a high degree of confidence, whereas less than 0.1% (ie, 2/4,399) were categorized with a low degree of confidence (Tables 2 and 3). The MALDI Biotyper CA System generated a “No Identification” result for 40 (0.9%) isolates (40/4,399) (Table 3).

Of the 2,262 aerobic gram-positive bacteria tested, the MALDI Biotyper CA System correctly identified 2,230 (98.6%) of these bacteria to the species level. In total, 2,185 (96.5%) of all aerobic gram-positive isolates were identified to the correct species level with a high level of confidence, with a matching log score of more than 2.0, whereas only 45 (2.0%) were correctly identified to the species level with a low level of confidence, with a matching log score between 1.70 and 1.99. Only 11 (0.5%) of the aerobic gram-positive bacteria were identified correctly to the genus level but assigned an incorrect species; of these, six were identified with a high level of confidence, whereas five were identified with a low level of confidence. Only two (<0.01%) of the aerobic gram-positive isolates tested were incorrectly identified to the genus level. Finally, the result of “No Identification” was given for 19 (0.84%) of the aerobic gram-positive isolates tested.

Of the 792 aerobic gram-negative bacteria tested, the MALDI Biotyper CA System correctly identified 780 (98.5%) isolates to the species level. Of these isolates, 767 (96.8%) were identified with a high level of confidence, whereas 13 (1.6%) were correctly identified to the species level with a low level of confidence. Ten (1.2%) of these isolates were correctly identified to the genus level but misidentified at the species level; nine (1.1%) were identified with a high level of confidence, whereas one (0.1%) was identified to this level but with a low level of confidence. Only two (0.2%) aerobic gram-negative isolates were misidentified at the genus level. None of the aerobic gram-negative bacteria were categorized as “No Identification.”

Of the 530 anaerobes tested, 522 (98.5%) were correctly identified to the species level; 516 (97.4%) were identified correctly to the species level or group with a high degree of confidence, whereas six (1.1%) isolates were correctly categorized to the species level but with a low degree of confidence. Only one (0.2%) of the anaerobes was correctly characterized to the genus level but received an incorrect species designation and with a low level of confidence. Three (0.6%) isolates were misidentified at the genus level, all of which were identified with a high level of confidence. The MALDI Biotyper CA System was unable to generate an identification for four (0.8%) of the anaerobic isolates.

The MALDI Biotyper CA System performed extremely well for the identification of yeast isolates at the species level (Table 2). In total, 97.8% (797/815) of yeast isolates were correctly identified to the species level; 90.9% (741/815) of these were categorized correctly to the species level with a high degree of confidence, whereas 6.9% (56/815) of these were correctly identified to the species level, albeit with a low degree of confidence. None of the yeast isolates were identified to the correct genus but incorrect species. Only one (0.1%) isolate was incorrectly identified (Table 3). The MALDI Biotyper CA System was unable to generate an identification (ie, “No Identification”) for 17 (2.1%) of the yeast isolates (Table 3).

Comparisons between the MALDI scores generated and the processing method for bacteria and yeasts are presented Table 4. Table 4 includes all measurements for the isolates (3,584 bacteria and 815 yeast) as well as all performed replicate testings (3,123 bacteria and 679 yeasts). The DT method alone identified 86.8% of the bacterial isolates with a high degree of confidence. The eDT method alone identified 89.9% of the bacterial isolates to the species level with a high degree of confidence. However, when combined results from both DT and eDT methods were used, 94.7% of the bacterial isolates generated a high confidence score of 2.0 or more for the species-level identification. The extraction method was reserved for difficult to identify isolates (ie, isolates with a DT or eDT score <2.0), which consisted of 1,232 of the isolates tested. In this schema, extraction considered separately was able to resolve (ie, generate a high confidence identification ≥2.0) for 992 (80.5%) isolates. However, when combined results from both DT and extraction methods were used, 97.0% of the bacterial isolates generated a high confidence score of more than 2.0 for the species-level identification.

References

MALDI-TOF mass spectrometry can be accurately described as a disruptive technology in clinical microbiology.¹¹ This technology, although relatively recently introduced, has revolutionized the way that bacteria and yeasts are identified and is being investigated for a variety of other microorganisms. Identifications that once took 1 or more days because of the necessity of subculture and biochemical test interpretation can now be accomplished within minutes at an extremely low cost.²⁰

The evaluation of the large number of organisms in this clinical trial disclosed much about the degree of confidence clinical microbiologists may have in the identification of particular organisms. It also revealed gaps in the current library and the microorganism groups that are clustered and not well differentiated by the described mass spectrometry method. Gaps in the microorganism coverage in the current library can, in part, be addressed through library updates that include the addition of a larger number of well-characterized isolates. Overall, the performance of the Bruker MALDI Biotyper CA System was outstanding; 98.6% of the aerobic gram-positive bacteria were correctly identified to the species level, with 96.6% being identified with a high level of confidence. Among the aerobic, gram-positive bacteria, this study included extensive coverage of the Corynebacterium species, which is a genus that contains many clinically relevant species that are challenging to identify to the species level with confidence using traditional methods. We studied 256 isolates from 17 Corynebacterium species or species complexes. All of these were correctly identified to the genus and species or species complex level (Table 3), with the exception of four of the 29 isolates of the Corynebacterium aurimucosum group for which a result of “No Identification” was generated. Suwantarat et al²¹ reported similar findings when they evaluated the ability of MALDI-TOF to accurately identify Corynebacterium species. They studied 231 isolates that represented 19 species and demonstrated that MALDI-TOF mass spectrometry was able to identify 99.6% to the genus level and 88.7% to the species level. MALDI-TOF mass spectrometry clearly represents an advance in the identification of Corynebacterium species.

One of the gaps identified by this trial was the inability to accurately identify Aerococcus viridans. The Bruker MALDI Biotyper CA System library used for this study did not sufficiently cover the MALDI-dependent diversity present among A viridans isolates, so some of the isolates studied failed to produce a high identification score. The addition of more strains of A viridans, with sufficient MALDI-dependent coverage of the variability present within this species, will be addressed in a future study submission and is expected to remedy this issue. This ability to add more isolates belonging to a particular species is a strength of the Bruker MALDI Biotyper CA System database concept. The addition of more isolates of a particular species may be all that is needed to address identification gaps in some instances, which does not influence the reliability of the rest of the database, which remains constant. Based on these findings, further isolates of A viridans will be tested and added to the database based on future study.

Conversely, there are other situations wherein there is insufficient MALDI-TOF spectra content variability to differentiate microorganisms within a parent group. As demonstrated in this study, Bacteroides xylanisolvens could not be differentiated by MALDI technology from other members of the Bacteroides ovatus group. The same is true for Bacteroides faecis and the Bacteroides thetaiotaomicron group, Bacteroides dorei and the Bacteroides vulgatus group, and members of the Elizabethkingia meningoseptica group, containing E meningoseptica, Elizabethkingia anophelis, and Elizabethkingia miricola. All these findings were confirmed by thorough visual analysis of the MALDI-TOF raw spectra. Clinical microbiologists should be aware of these types of groupings so that additional studies can be performed if differentiation is needed for clinical purposes.

One of the significant but known limitations of MALDI-TOF mass spectrometry is the differentiation of S pneumoniae from the S mitis/S oralis group.²² This study included 30 S pneumoniae isolates, of which all were correctly identified to the species level with a high confidence score using a combination of DT, eDT, and extraction. Only two of 73 S mitis/oralis group isolates were misidentified as S pneumoniae at the species level using the DT or eDT method. However, after extraction, all S mitis/oralis group isolates were identified correctly. In addition, based on these results, all S pneumoniae isolates were retrospectively reanalyzed using the extraction method to define the spectrum profile that would afford the correct identification for all preparation methods (ie, DT, eDT, and extraction). The processing-to-processing comparisons disclosed the importance of a full extraction in differentiating these organisms. In fact, when this differentiation is not achieved using the DT or eDT processing method, a software alert has been included that recommends the use of extraction if further differentiation is desired.

In other instances, MALDI-TOF mass spectrometry readily differentiates microorganisms that have been traditionally included in a group or may be challenging to differentiate by traditional methods. For example, the Candida parapsilosis group, which is usually simply identified as C parapsilosis, actually consists of three organisms: C parapsilosis, Candida metapsilosis, and Candida orthopsilosis. These yeasts are readily differentiated by the MALDI Biotyper CA System, which raises a number of issues. Foremost, it is important that clinicians understand that these “new” organisms, if reported as C metapsilosis or C orthopsilosis, are in fact part of the same C parapsilosis complex to which they are accustomed. This differentiation to the species level, however, may be clinically important, as some have suggested differences in the antifungal susceptibility profiles of these closely related fungi.²³ Another example of species-level identification at a high confidence level is the differentiation of Cryptococcus neoformans and Cryptococcus gattii. The data in this trial demonstrated the ability of the MALDI Biotyper CA System to accurately and reliably differentiate two variants of C neoformans from C gattii, for which there are important clinical and epidemiologic differences.

The processing comparisons undertaken in this study provide insights regarding the efficacy of each extraction method used independently, as well as in select combinations (Table 4). Combined extraction approaches could be performed concurrently or in tandem, depending on laboratory preference. The DT method, which consists of the fewest steps, demonstrated excellent performance in the ability to characterize the majority (86.8%) of bacterial isolates with a high degree of confidence (ie, a MALDI score of ≥2.0). This suggests that most isolates can be readily characterized with a high degree of confidence with minimal steps. If the MALDI score is lower than desired for a particular isolate, then the simple eDT technique may result in a score of 2.0 or more. Extraction could be reserved for the most challenging identifications that fail to be classified correctly by the DT and/or eDT methods.

In contrast to bacteria, the DT method alone was significantly inferior to the eDT method alone for the identification of yeast isolates with a high degree of confidence (48.9% vs 74.0%, respectively). The combination of DT and eDT methods produced the best result, with 81.6% of the isolates tested generating a high confidence score of 2.0 or more. As above, the extraction method was reserved for those isolates that generated a score of less than 2.0 with either the DT or eDT methods, which consisted of 901 isolates. After extraction, 606 (67.3%) of these isolates were resolved (ie, generated an identification ≥2.0). When the results from both the DT and extraction methods were combined, 83.4% of the yeast isolates could be categorized with a high degree of confidence and a further 10.7% with a low degree of confidence, giving a combined overall performance of 94.1% for yeast. An implication of these findings is the primary use of the eDT method for yeasts, with the reservation of the alternate processing methods if an acceptable identification is not achieved.

In conclusion, the MALDI Biotyper CA System performed well in this multicenter FDA submission trial and accurately characterized most clinically important bacteria and yeasts with which it was challenged. This 510(k) submission was approved by the FDA. MALDI-TOF mass spectrometry represents a significant advance in the ability of microbiologists to achieve rapid and accurate identification of medically important bacteria and fungi. As these products are more widely used, the performance characteristics can be more thoroughly studied and gaps addressed either through the inclusion of more strains in future library updates or by simply defining which groups can or cannot be sufficiently differentiated by this technology. Although these applications are groundbreaking, they are only the beginning, as other applications under investigation include the identification of mycobacteria and aerobic actinomycetes, filamentous fungi, detection of resistance to certain antimicrobial agents, strain typing, and more.^24‐31

In conclusion, the MALDI Biotyper CA System performed well in this multicenter FDA submission trial and accurately characterized most clinically important bacteria and yeasts with which it was challenged. This 510(k) submission was approved by the FDA. MALDI-TOF mass spectrometry represents a significant advance in the ability of microbiologists to achieve rapid and accurate identification of medically important bacteria and fungi. As these products are more widely used, the performance characteristics can be more thoroughly studied and gaps addressed either through the inclusion of more strains in future library updates or by simply defining which groups can or cannot be sufficiently differentiated by this technology. Although these applications are groundbreaking, they are only the beginning, as other applications under investigation include the identification of mycobacteria and aerobic actinomycetes, filamentous fungi, detection of resistance to certain antimicrobial agents, strain typing, and more.^24‐31