European Journal of Clinical Microbiology & Infectious Diseases

, Volume 31, Issue 9, pp 2257–2262

Identification of clinical isolates of anaerobic bacteria using matrix-assisted laser desorption ionization-time of flight mass spectrometry

Authors

    • Department of Laboratory Medicine, Clinical CenterNationals Institutes of Health, Microbiology Service
  • S. K. Drake
    • Critical Care Medicine Department, Clinical CenterNationals Institutes of Health
  • F. Stock
    • Department of Laboratory Medicine, Clinical CenterNationals Institutes of Health, Microbiology Service
  • P. R. Murray
    • Department of Laboratory Medicine, Clinical CenterNationals Institutes of Health, Microbiology Service
    • BD Diagnostics
Article

DOI: 10.1007/s10096-012-1563-4

Cite this article as:
Fedorko, D.P., Drake, S.K., Stock, F. et al. Eur J Clin Microbiol Infect Dis (2012) 31: 2257. doi:10.1007/s10096-012-1563-4

Abstract

We evaluated the use of matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) for the rapid identification of anaerobic bacteria that had been isolated from clinical specimens and previously identified by 16s rRNA sequencing. The Bruker Microflex MALDI-TOF instrument with the Biotyper Software was used. We tested 152 isolates of anaerobic bacteria from 24 different genera and 75 different species. A total of 125 isolates (82%) had Biotyper software scores greater than 2.0 and the correct identification to genus and species was made by MALDI-TOF for 120 (79%) of isolates. Of the 12 isolates with a score between 1.8 and 2.0, 2 (17%) organisms were incorrectly identified by MALDI-TOF. Only 15 (10%) isolates had a score less than 1.8 and MALDI-TOF gave the wrong genus and species for four isolates, the correct genus for two isolates, and the correct genus and species for nine isolates. Therefore, we found the Bruker MALDI-TOF MicroFlex LT with an expanded database and the use of bacteria extracts rather than whole organisms correctly identified 130 of 152 (86%) isolates to genus and species when the cut-off for an acceptable identification was a spectrum score ≥1.8.

Introduction

The identification of bacteria and yeasts has traditionally been based on phenotypic characteristics obtained by Gram’s stain, growth and colonial morphology on culture media, and biochemical characteristics determined by the ability to use substrates either during growth or in the presence of preformed enzymes. Rapid and automated systems have been approved by regulatory agencies such as the United States Food and Drug Administration (FDA) for identification of bacteria and yeasts isolated from clinical specimens. Many clinical microbiology laboratories still use these methods to identify clinical isolates because they are well established and inexpensive, but they are known to be inaccurate or inconclusive especially with uncommon or fastidious organisms. The current gold standard for identification of clinically important microorganisms relies on PCR amplification and sequencing of the 16sRNA gene for bacteria and the internal transcribed spacer (ITS) region for yeasts. There are no FDA-approved commercial systems for this purpose, it requires technical expertise and expensive equipment, and is therefore not available to many clinical laboratories.

Recent advancements in mass spectrometry (MS) have allowed for the rapid and accurate identification of bacteria and fungi using whole cells or cell lysates. Two commercial companies (Bruker Daltonik, Bremen, Germany and Shimadzu Corporation, Kyoto, Japan) produce user-friendly matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) MS systems with software and databases for the identification of microorganisms isolated from clinical specimens [1, 2]. These systems have been used to identify aerobic and anaerobic bacteria [37], mycobacteria [8, 9], nocardia [10], and yeasts [6] isolated on solid media from clinical specimens. MALDI-TOF has also been recently applied to the identification of bacteria and yeast directly from positive blood culture bottles [3, 11, 12]. In this study we evaluated the Bruker MALDI-TOF MS system for the identification of clinical isolates of anaerobic bacteria that had been characterized by 16sRNA gene sequencing.

Materials and methods

Type and reference strains and clinical isolates

Type and reference strains from the American Type Culture Collection (ATCC) were used in an initial validation of the method and database. Clinical isolates were recovered from frozen stock of organisms that had been isolated from patient specimens at the NIH Clinical Center. The distribution of organisms tested by MALDI-TOF reflects the distribution of organisms in our frozen stock of clinical isolates. All anaerobic bacteria were cultivated on CDC Blood Agar medium (Remel, Lenexa, KS). Four type strains and one reference strain of the Bacteroides fragilis group of organisms (B. fragilis ATCC 23745, B. fragilis ATCC 25285, B. thetaiotaomicron ATCC 29741, B. uniformis, P. distasonis ATCC 8503) were also grown on pre-reduced Bacteroides bile esculin (BBE) agar and Brucella blood agar (Anaerobe Systems, Morgan Hill, CA) for 24 hours to test for the influence of culture media on the spectra generated in the MALDI-TOF instrument. All clinical isolates were identified by 16S rDNA sequencing using the MicroSeq 16S rDNA Bacterial Identification Kit (Applied Biosystems, Carlsbad, CA) and a 3100 Genetic Analyzer (Applied Biosystems). DNA sequences were processed using LaserGene (DNASTAR, Inc., Madison, WI). A BLAST search was performed using GenBank and the University of Michigan Database Project and identifications were made by comparing our sequences to sequences of type strains in the databases using a  ≥99% match with either a 500 bp sequence or a 1500 bp sequence from GenBank (http://www.ncbi.nlm.nih.gov/) and the University of Michigan Database Project (http://rdp.cme.msu.edu/).

MALDI-TOF

One large colony or multiple small colonies (enough to fill about one half of a 10-μl inoculating loop) of a bacterial isolate to be tested was suspended in 70% ethanol in a 1.5-ml microcentrofuge tube. Extraction of bacteria, matrix preparation, spotting of the steel target plate, and calibration of the instrument were performed as previously described [12]. A Bruker MALDI-TOF MicroFlex LT mass spectrometer was used to generate spectra from the bacterial extracts, and the Biotyper software (Version 2.0.4) was used to analyze the results.

Results

Type and reference strains

All of the type and reference strains listed in Table 1 were correctly identified to the species level by the MicroFlex LT mass spectrophotometer with a Biotyper software score value greater than 2.0. Sixteen of these organisms were type strains that are included in the Bruker database. We routinely use Bacteroides Bile Esculin (BBE) agar as the only selective medium in our anaerobe culture set-up. We tested the effect of culture medium on MALDI-TOF scores using Bacterioides fragils group organisms grown on BBE, Brucella Agar, and CDC Agar. There was no effect of culture medium on the ability of the MALDI-TOF to correctly identify the members of the Bacteroides fragilis group that were tested. All spectrum scores were greater than 2.0.
Table 1

Type and reference strains used to validate the method and Biotyper software

Organism name

ATCC Number

Gram positive rods

 Clostridium septicum

12464

 Clostridium sporogenes

3584

 Clostridium histolyticum

19401

 Clostridium sordellii

9714

 Clostridium sporogenes

11437

 Clostridium perfringens

13124

 Lactobacillus casei ssp casei

393

 Actinomyces naeslundii

27038

 Actinomyces odontolyticus

17929

 Propionibacterium acnes

11827

Cocci

 Finegoldia magna

29328

 Peptostreptococcus anaerobius

27337

 Veillonella parvula

10790

Gram negative rods

 

 Bacteroides fragilis

23745

 Bacteroides fragilis

25285

 Bacteroides ovatus

8483

 Bacteroides uniformis

8492

 Bacteroides thetaiotaomicron

29741

 Parabacteroides distasonis

8503

 Prevotella melaninogenica

25845

 Fusobacterium necrophorum ssp necrophorum

25286

 Fusobacterium nucleatum ssp nucleatum

25586

Clinical isolates

A total of 152 clinical isolates were identified by 16SrDNA sequencing and tested by MALDI-TOF. One hundred twenty-five isolates (82%) had Biotyper software scores greater than 2.0 (Tables 2, 3 and 4). Correct identification to genus and species was made by MALDI-TOF for 120 (79%) of the 152 clinical isolates. This was not statistically different from the 16SrDNA gold standard (p = 0.789). The remaining five isolates were correctly identified to the genus level (4 isolates of Fusobacterium nucleatum and 1 isolate of Actinomyces graeventitzii). Twenty-seven isolates (18%) had a MALDI score less than 2.0. Of the 12 isolates with a score between 1.8 and 2.0, 10 (83%) were correctly identified by MALDI-TOF. An isolate of Anaerococcus hydrogenalis was incorrectly identified as Peptoniphilus harei with a score of 1.855, and an isolate of Anaerococcus vaginalis was identified as Anaerococcus hydrogenalis with a score of 1.941 (Table 3). Only 15 (10%) isolates had a score less than 1.8, and MALDI-TOF gave the wrong genus and species for four isolates, the correct genus for two isolates, and the correct genus and species for nine isolates. Therefore, using a score of ≥1.8 as a cut-off, 130 of 152 (86%) clinical isolates were correctly identified to genus and species (p = 0.895), and 136 of 152 (89%) were identified to the correct Genus.
Table 2

Anaerobic gram positive rods

Genus

Species

MALDI spectrum score

<1.8

1.8–2.0

>2.0

Clostridium

aldenense

  

1

bifermentans

  

2

butyricum

  

2

cadaveris

  

3

citroniae

  

1

clostridiforme

  

1

difficile

  

2

glycolicum

  

1

hathewayi

  

1

innocuum

 

1

3

paraputrificum

  

1

perfringens

  

4

ramosum

  

4

septicum

  

2

symbiosum

  

1

tertium

  

2

Actinomyces

graevenitzii

  

1 (genus)a

israelii

 

1

 

meyeri

1

  

odontolyticus

 

2

1

naeslundii

1 (mis-ID)b

  

neuii

 

1

 

turicensis

  

1

ureogenitalis

  

2

viscosus

1

  

Bifidobacterium

breve

  

1

dentium

1

 

1

longum

  

1

Capnocytophaga

gingivalis

1 (mis-ID)b

  

sputigena

  

3

Eggerthella

lenta

  

2

Eubacterium

limosum

  

1

Lactobacillus

gasseri

  

1

paracasei

  

1

planatarum

  

1

rhamnosus

  

1

vaginalis

  

1

zeae

  

1

Atopobium

rimae

 

1

 

Leptotrichia

buccalis

 

1

 

Propionibacterium

acnes

  

4

granulosum

  

2

Totals

 

5

7

57

Identification to correct genus only

b Incorrect identification of both genus and species

Table 3

Anaerobic cocci

Genus

Species

MALDI spectrum score

<1.8

1.8–2.0

>2.0

Anaerococcus

murdochii

  

2

octavius

  

1

provetii

2

  

hydrogenalis

1

1 (mis-ID)

 

vaginalis

1 (genus)a

1 (genus)

 

Finegoldia

magna

  

4

Gemella

haemolysans

  

1

Parvimonas

micra

 

1

3

Peptonephilus

harei

  

6

Peptostreptococcus

anaerobius

 

1

3

Staphylococcus

saccharolyticus

  

1

Veillonella

atypica

  

1

parvula

 

1

4

Totals

 

4

5

26

Identification to correct genus only

b Incorrect identification of both genus and species

Table 4

Anaerobic gram negative rods

Genus

Species

MALDI spectrum score

<1.8

1.8–2.0

>2.0

Bacteroides

fragilis

  

6

ovatus

  

5

thetaiotaomicron

  

3

uniformis

  

2

ureolyticus

  

1

vulgatus

  

5

Fusobacterium

nucleatum

1 (mis-ID)

 

5 (4 genus)

gonidiaformans

  

2

necrophorum

1

  

mortiferum

1 (mis-ID)

  

varium

2 (1 genus)

  

Prevotella

bergensis

  

1

bivia

  

4

buccae

  

1

disiens

  

1

melaninogenica

1

  

nanceiensis

  

2

Parabacteroides

distasonis

  

2

Porphyromonas

gingivalis

  

1

Slakia

exigua

  

1

Totals

 

6

0

42

Identification to correct genus only

b Incorrect identification of both genus and species

Table 2 shows results from testing gram-positive rods. Isolates in the genus Actinomyces were the least likely to give high spectrum scores. We tested 11 isolates of nine different species of Actinomyces. All isolates were properly identified to genus and species, but only 4 (36%) isolates had a score of greater than 2.0 and eight (73%) had a score of greater than 1.8. Of the three isolates with scores below 1.8, A. viscosus had one entry in the Bruker database and A. meyeri and A. naeslundii both had two entries in the database.

Table 4 shows results from testing gram-negative rods. Isolates in the genus Fusobacterium created the most incorrect identifications. The MALDI-TOF was challenged with the following subspecies of Fusobacterium nucleatum: F. nucleatum ssp fusiforme, F. nucleatum ssp vincetii, and F. nucleatum ssp nucleatum. All four F. nucleatum ssp fusiform were identified as Fusobacterium naviforme by the MALDI-TOF with scores greater than 2.0. The isolate of F. nucleatum ssp vincentii was identified as Lactobacillus paracasei with a score of 1.3. The MALDI-TOF database contained a single entry for each of these organisms. The F. nucleatum subspecies in the database were type strains, but the F. naviforme was not a type strain. The two isolates of Fusobacterium gonidiaformans were correctly identified with a score greater than 2.0, but the remaining three species produced spectrum scores below 1.8.

There were 16 isolates representing ten different species that were not in the Bruker database when the study began. All had scores below 1.8 and incorrect genus and species as first choices for the identification except for an isolate of Capnocytophaga sputigena (score 1.909 Capnocytophaga ochraceae). Near the end of our study an updated database containing 3,996 entries was received from Bruker. The updated database was compared to the previous database and of the ten species not identified using the previous database because they were not included; the following six species were included in the updated database: Clostridium glycolyticum, Clostridium symbiosum, Actinomyces israeli, Eggerthella lenta, Eubacterium limosum, and Capnocytopgaga sputigena, and the number of P. acnes strains in the database increased from five to nine. Following the criteria described previously by Saleeb et al. [9] we added entries of the following ATCC type strains to the database: Clostridium glycolicum ATCC 14880, Eggerthella lenta ATCC 25559, Leptotrichia buccalis ATCC 14201, Capnocytophaga sputigena ATCC 33612, Bacteroides ureolyticus ATCC 33387, Fusobacterium gonidiaformans ATCC 25563, and Atopobium rimae ATCC 49626. The 16 organisms were retested and all were correctly identified. All but L. buccalis (score 1.877) and A. israelii (score 1.933) gave scores greater than 2.0. All organisms with scores below 1.8 were re-evaluated using the new database and 11 organisms were correctly identified with a score ≥2.0 and the total was 12 when a cutoff of ≥1.8 was used.

Discussion

In this study the Bruker MALDI-TOF MicroFlex LT mass spectrometer with the Biotyper software (version 2.0.4) allowed identification of anaerobic bacteria isolated from clinical specimens with a high degree of precision if the organism was in the database. The Biotyper software compares the spectrum of the unknown isolate to the peak-list based entries of bacterial strains in the database and provides the user with spectra scores (similarity scores) and closest matches to organisms in the database. A spectrum score less than 1.8 is considered an unreliable identification by the Biotyper software. For isolates of anaerobic bacteria from patient specimens we found the Bruker MALDI-TOF MicroFlex LT correctly identified 130 of 152 (86%) isolates to genus and species when the spectrum score was 1.8 and above, which was used as a cut-off point for an acceptable identification. This is higher than the identification rates for anaerobic bacteria of 61–67% previously reported in the literature [2, 4, 7]. There are two explanations for the enhanced performance of the Bruker MALDI-TOF for anaerobe identification in our study. First, it is well described in the literature that the performance of a MALDI-TOF system for identification of any microorganism directly correlates with the quality of the database [2, 4, 13]. We used the most recent database from Bruker that contained 3,996 entries which we had expanded. Veloo et al. [2] compared the Bruker system with the Shimadzu system and found the Shimadzu system gave more correct identifications of anaerobic bacteria because it had a larger database. They reported a remarkable improvement of correct identifications (from 51% to 63.3%) when they utilized an updated database (3,476 entries vs 3,740 entries) from Bruker to evaluate their data. Second, we used bacterial extracts rather than deposition of whole organisms directly onto the target plate. La Scola et al. [4] reported an identification rate of only 61% for anaerobes, and Justesen et al. [7] reported a rate of 67.2% using a Bruker MALDI-TOF system with whole organisms applied to the target plate. Veloo et al. [2] compared extracts to whole organism identifications in the Bruker Microflex and they reported improved identification rates when extracted organisms were used. They indicate that the Bruker database was created entirely using extracts of organisms. They also say that they feel pretreatment is most helpful for gram-positive organisms. Alatoom et al. [14] compared whole organism identification to extracts for the identification of aerobic gram-positive cocci in the Bruker system. They reported identification rates of 56% to genus and 20% to species when using whole organisms and 95% to genus and 69% to species when using extracts. It appears that bacterial extracts are superior to whole organism deposition for the identification of bacteria when using MALDI-TOF [2, 14]. Although the use of extracts makes the process slightly more complicated and slightly longer, we expect the increased sensitivity would reduce the number of isolates that need to be retested due to a low score.

The only selective medium we use for primary culture of anaerobes is BBE medium. We found no difference in MALDI-TOF scores (all > 2.0) for four common members of the B. fragilis group when colonies were taken from growth on BBE, CDC blood agar, and Brucella blood agar. Although it is possible that other members of the B. fragilis group may not show this same response, Veloo et al. previously reported that they found no effect of culture medium when a subset of 58 anaerobes was grown on both Trypticase Soy agar with sheep blood and Brucella agar with sheep blood and subsequently tested with a Bruker Microflex system [2].

Using colonies of organisms grown on solid media Seng et al. [13] were able to identify 50 of 58 (86%) P. acnes isolates to genus and species, La Scola et al. [4] identified 97 of 130 (75%), and Veloo et al. [2] identified three of six (50%) P. acnes using a Bruker MALDI-TOF instrument. Veloo et al. [2] indicated that the Bruker database they used had entries from five different P. acnes strains, and they point out that differences in the number of entries in the database has a direct correlation with the ability of a MALDI-TOF system to produce a correct identification. All four of our P. acnes isolates were correctly identified with a spectrum score greater than 2.0 (Table 2). At the beginning of our study the database contained entries representing five different strains of P. acnes, but the updated database contained nine different entries. An expanded database would clearly enhance the ability of the software to make a correct identification for organisms whose number of entries has been increased [2].

In our study MALDI-TOF had the most difficulty in identifying organisms in the genus Fusobacterium to the species level (Table 4). The updated database we used contained 13 entries for ten different species or subspecies plus an organism identified as Fusobacterium sp. Each species or subspecies had a single entry in the database with the exception of Fusobacterium canifelinum and F. necrophorum ssp necrophorum, each of which had two entries in the database. In our study only four of 12 (33%) Fusobacterium species were identified to the genus level with a score above 1.8, only three (25%) isolates were identified to the species level with a score above 1.8, and all four F. nucleatum ssp fusiform were identified as Fusobacterium naviforme by the MALDI-TOF with scores greater than 2.0 (Table 4). Conrads et al. [15] used 16S-23S rDNA internal transcribed spacer sequences to analyze phylogenetic relationships among species of Fusobacterium and they reported that F. nucleatum ssp. fusiforme and F. naviforme are very closely related to one another. Clearly, the database for Fusobacterium needs to be expanded to enhance the ability of the Bruker MALDI-TOF Microflex to correctly identify isolates of Fusobacterium to the species level.

The present study demonstrates that the Bruker Microflex MALDI-TOF instrument with Biotyper Software and the most recent database compares favorably to 16S rRNA gene sequencing for the identification of anaerobic bacteria isolated from clinical specimens. Because we used extracts of organisms instead of whole organisms in our evaluation of the Bruker system we feel our evaluation provides the most accurate assessment of the Bruker system since it’s database is composed entirely of spectra generated from organism extracts [2]. Identification of anaerobic bacteria using MALDI-TOF is rapid and inexpensive and can have a great impact on the work-up of bacteria isolated in anaerobe cultures. With MALDI-TOF microbiologists no longer have to perform presumptive identifications using tests such as special potency antimicrobial disks, sodium polyanethol sulfonate (SPS) disks, nitrate disks, or lecitinase and lipase reactions that require growth of suspect anaerobes. Rapid tests such as catalase or spot indole are also unnecessary. If there is enough growth with isolated colonies, identification of aerobes and anaerobes growing from primary culture plates of an anaerobic culture can be rapidly and accurately identified by MALDI-TOF without subculture to demonstrate a bacterium’s atmospheric requirements or restrictions for growth.

Acknowledgement

This research was supported by the Intramural Research Program of the NIH Clinical Center, Department of Laboratory Medicine.

Copyright information

© US Government 2012