Goslar, 09/10/2007 Identification of Microorganisms using MALDI-TOF MS profiling: Adopted sample preparation methods and bioinformatic approaches Dr.

Goslar, 09/10/2007

Identification of Microorganisms using MALDI-TOF MS profiling:Adopted sample preparation methods and bioinformatic approaches Dr. Markus KostrzewaBruker Daltonik GmbH, Leipzig

Goslar, 09/10/2007

MALDI-TOF MS microorganism identification

Identified species

Select a colony

Prepare ontoa MALDI target plate

Unknownmicrorganism

?

Data interpretation

Generate MALDI-TOFprofile spectrum

Goslar, 09/10/2007

Target/Acceleration

Time-of-Flight Molecular MassDesorption/Ionisation

DetectorDrift Region

m/z

a.i.

Mass Spectra

Laser

MALDI-TOF mass spectrometry

Goslar, 09/10/2007

Sample preparation

• Direct “cell smear“ methodmost simple method, applicable to many bacteria

• Organic solvent extractionimproved quality for difficult bacteria, yeast, fungi

• Mechanical cell disruption (e.g. sonication)In case of very ridgid cell walls

Compatibility of different procedures

Protocols for inactivation and shipment of microorganisms are available

Goslar, 09/10/2007

E.coli4

36

4.0

6

53

80

.64

62

54

.64

63

15

.49

50

96

.01

71

57

.65

72

73

.87

64

10

.90

78

70

.62

83

68

.99

0

1000

2000

3000

4000

5000Inte

ns.

[a.u

.]

4000 4500 5000 5500 6000 6500 7000 7500 8000

m/z

ribosomal Protein m/zRL36 4364,33RS32 5095,82RL34 5380,39RL33meth. 6255,39RL32 6315,19RL30 6410,60RL35 7157,74RL29 7273,45RL31 7871,06RS21 8368,76

MALDI-TOF MS profile spectrumPositive linear modeMass range 2-20 kDa

Goslar, 09/10/2007

Improved quality by adopted sample preparation

Pichia jadiniiextraction method

0.00

0.25

0.50

0.75

1.00

1.25

1.50

4x10

Inte

ns.

[a.u

.]

Pichia jadiniidirect smear method

0.00

0.25

0.50

0.75

1.00

1.25

4x10

Inte

ns.

[a.u

.]

2000 4000 6000 8000 10000 12000 14000 16000 18000

m/z

Identification score

2.445

1.997

Goslar, 09/10/2007

Psdm. oleovorans B396_Medium 360

0

1000

Psdm. oleovorans B396_Medium 464

0

1000

Psdm. oleovorans B396_Medium 53

0

1000

Psdm. oleovorans B396_Medium 65

0

1000

Psdm. oleovorans B396_Medium 98

0500

1000

Psdm. oleovorans B396_MRS10

010002000

Psdm. oleovorans B396_YPD

010002000

4000 5000 6000 7000 8000 9000 10000 11000m/z

Pseudomonas oleovorans grown on different media

Low influence of culture conditions

Goslar, 09/10/2007

Arthrobacter,effect of storage products

Taken from:Vargha M et al.J Microbiol Methods. 2006

Possible influence of growth state

16h

24h

64h

0

1000

2000

3000

4000

3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 m/z

Clostridium butyricum,effect of sporulation

Coop. with Prof. Krüger, Dr. Grosse-Herrenthey, Leipzig, Germany

Goslar, 09/10/2007

MALDI BioTyper 2.0 - GUI

Unknown samples

Match against microbial reference database

Identification result

Goslar, 09/10/2007

MALDI BioTyper 2.0 – realtime analysis

Wizard guiding through setup from measurement to data analysis

BioTyper Automation Control Wizard

Define Project

Analyte Placement

Select Methods

Start

MALDI BioTyper 2.0: realtime Analysis

• Start Automation Control Wizard.• Use a SOP and do not bother with

instrument settings.• Results available directly after measurement

Goslar, 09/10/2007

MALDI BioTyper – Algorithms

Pattern matchingweighted pattern matching

Principle component analysisCluster analysis Correlation analysis

Goslar, 09/10/2007

Score based pattern matching

Calculation of a matching score based on:

Rel Score% matches of the reference spectrum

(e.g. 6 / 10 = 0.6)

Rel P-Num.% matches of the unknown spectrum

(e.g. 6 / 20 = 0.3)

I-Corr.value of intensity correlation

Unknown microorganism is matched against each Main spectrum in a library. Ranking according to matching score, threshold for IDRobust standard method for species ID

Goslar, 09/10/2007

Pseudo-Gel view

M/Z [Da]

File

no.

4000 5000 6000 7000 8000 9000 10000 11000 12000

1 2 3 4 5 6 7 8 910111213141516171819202122232425262728293031323334353637383940

Neisseria meningitidis serotypes

A

W135

X

Y

How about subtyping?

Pseudo-Gel view

M/Z [Da]

File

no.

6100 6200 6300 6400 6500 6600 6700 6800 6900 7000 7100

1 2 3 4 5 6 7 8 910111213141516171819202122232425262728293031323334353637383940

Goslar, 09/10/2007

Incorrect hirachical clustering of three Neisseria meningitidis serogroups after PCA

-2

0

2

4

-2-1

01

2-1

-0.5

0

0.5

1

1.5

PC1

3D scatter plot - hierarchical clustering

PC2

PC

3

Principle component analysis

PCA is looking for the largest variations in a given group.If measurement variations are larger than inter-species/ subspecies variations it may fail! Depending strongly on standardization of measurement!

Goslar, 09/10/2007

Weighted pattern matching

• Batch weighting:Automated generation of a weighted main spectra library; every main spectrum of a library is compared with all the other main spectra

• Manual weighting:Weight of each peak in a main spectrum can be edited manually

Combination of both procedures is possible

Hierachical approach in combination with standard pattern matching

Characteristic peaks are selected and weighted by occurence in subgroups, intensity, and frequency

Goslar, 09/10/2007

Weighted pattern matching

Identification Results weighted

Detected Species log(Score)--------------------------------------------------------------Sp. 1 Serogruppe_A 2.677Ser.A Serogruppe_Y 2.150 Serogruppe_W135 2.044 Serogruppe_X 2.026Sp. 2 Serogruppe_W135 2.339Ser.W135 Serogruppe_Y 2.123 Serogruppe_X 1.784 Serogruppe_A 1.571Sp. 3 Serogruppe_X 2.665Ser.X Serogruppe_W135 2.033 Serogruppe_Y 1.902 Serogruppe_A 1.136Sp. 4 Serogruppe_Y 2.294Ser.Y Serogruppe_W135 2.126 Serogruppe_X 1.958 Serogruppe_A 1.617

Neisseria meningitidis serogroups

Correct identification of subspecies through weighting of specific peaks.

Expansion of pattern matching towards subspecies detection.

Goslar, 09/10/2007

Correlation analysis

Composite Correlation Index Map

Species

Spe

cies

2 4 6 8 10 12

2

4

6

8

10

12

Color code:dark red – highest correlationdark blue – lowest correlation

Correlation analysis of different Salmonella enterica serovars:Correlation analysis according to Arnold & Reilly, RCMS, 1998, modified

1. 1849_Hadar_VAB 2. 371_enteritidis_VAB 3. 042_typhimurium_O5_VAB 4. 104_enteritidis_VAB 5. 123_typhimurium_O5_VAB 6. 163_Virchow_VAB 7. 188_Dublin_VAB 8. 202_Infantis_VAB 9. 242_Infantis_VAB 10. 285_Virchow_VAB 11. 506_Hadar_VAB 12. 754_Agona_VAB

Goslar, 09/10/2007

Microorganism databases

Acetobacter aceti subsp. acetiAcetobacter pasteurianus subsp.lovaniensisAcetobacter pasteurianus subsp.pasteurianusActinomadura aurantiacaActinomadura libanoticaActinomadura lividaAgrobaterium tumefaciensArthrobacter globiformisArthrobacter oxydansArthrobacter pyridinolisArthrobacter sulfureusBacillus alcalophilusBacillus cohniiBacillus sphaericusBrevibacillus brevisBrevibacterium linensCellulomonas flavigenaCellulomonas turbataCorynebacterium glutamicumComamonas testosteroniiGluconobacter oxydans subsp. oxydansGluconobacter oxydans subsp.oxydansGordonia amaraeGordonia rubropertinctaGordonia terraeHalomonas denitrificansHalomonas elongataHalomonas elongataHalomonas halmophilaHalomonas halmophilaHydrogenophaga flavaHydrogenophaga pseudoflava

Methylobacterium mesophilicumMethylobacterium organophilumMethylobacterium radiotoleransMethylobacterium rhodesianumParacoccus versutusParacoccus versutusPseudomonas balearicaPseudomonas fluorescensPseudomonas fluorescensPseudomonas oleovoransPseudomonas putidaPseudomonas stutzeriPseudonocardia hydrocarbonoxydansRhizobium leguminosarumRhodococcus coprophilusRhodococcus fasciansRhodococcus globerulusRhodococcus rhodniiRhodococcus rhodochrousRhodococcus ruberSinorhizobium melilotiStarkaya novellaStarkaya novellaStreptomyces albusStreptomyces avidiniiStreptomyces azureusStreptomyces badiusStreptomyces griseusStreptomyces hirsutusStreptomyces lavendulaeStreptomyces phaeochromogenesStreptomyces violaceoruberStreptomyces viridisporus

Libraries:

• Generation of reference pattern for new microorganisms by users

• Ready-to-use libraries with microbial strains for direct identification

Goslar, 09/10/2007

• Minimal sample preparation

• Powerful bioinformatic approaches

• Species to strain resolution, mixture detection

• High throughput at low costs per analysis

• Non-expert identification possible

• Dedicated databases of high quality

Conclusions

Goslar, 09/10/2007

The BDAL BioTyper team:

Thomas MaierKristina SchlosserThomas WenzelThorsten MieruchStefan KlepelUwe RennerJan-Henner WurmbachKarl-Otto KräuterAlexander Rueegg

Thanks to:

… and many cooperation partners!

In particular:Prof. Stackebrandt,Dr. Schumann