MALDI-TOF and Tandem MS for Clinical Microbiology E-Book

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 1273

Veröffentlichungsjahr: 2017

Beschreibung

This book highlights the triumph of MALDI-TOF mass spectrometry over the past decade and provides insight into new and expanding technologies through a comprehensive range of short chapters that enable the reader to gauge their current status and how they may progress over the next decade.  This book serves as a platform to consolidate current strengths of the technology and highlight new frontiers in tandem MS/MS that are likely to eventually supersede MALDI-TOF MS. Chapters discuss:
Challenges of Identifying Mycobacterium to the Species level
Identification of Bacteroides and Other Clinically Relevant Anaerobes
Identification of Species in Mixed Microbial Populations
Detection of Resistance Mechanisms
Proteomics as a biomarker discovery and validation platform
Determination of Antimicrobial Resistance using Tandem Mass Spectrometry

Leseprobe

Table of Contents

Cover

Title Page

List of Contributors

Preface

Part I: MALDI‐TOF Mass Spectrometry

1 A Paradigm Shift from Research to Front‐Line Microbial Diagnostics in MALDI‐TOF and LC‐MS/MS:

1.1 Introduction

1.2 Overcoming the Variable Parameters of MALDI‐TOF MS Analysis: Publication of the First Database in 2004

1.3 SELDI‐TOF MS: A Powerful but Largely Unrecognized Microbiological MALDI‐TOF MS Platform

1.4 MALDI‐TOF MS as a Platform for DNA Sequencing

1.5 Insights into the Proteome of Major Pathogens 2005–2009: Field Testing of MALDI‐TOF MS

1.6 2010–2011: The Triumph of MALDI‐TOF MS and Emerging Interest in Tandem MS for Clinical Microbiology

1.7 Preparations for MALDI‐TOF MS Analysis on a Grand Scale: The Looming London 2012 Olympics

1.8 Investigating the Detection and Pathogenic Potential of E. coli O104:H4 during Outbreak of 2011

1.9 Conclusions

References

2 Criteria for Development of MALDI‐TOF Mass Spectral Database

2.1 Introduction

2.2 Commercially Available Databases

2.3 Establishment of User‐Defined Databases

2.4 Species Identification/Control of Reference Strains to Be Included into a Database

2.5 Sample Preparation

2.6 MALDI‐TOF MS Measurement

2.7 Quality Control during Creation and after Establishment of Reference Libraries

2.8 Common Influencing Factors for MALDI‐TOF MS

2.9 User‐Created and Shared Databases: Examples and Benefits

References

3 Applications of MALDI‐TOF Mass Spectrometry in Clinical Diagnostic Microbiology

3.1 Introduction

3.2 Principle of Microorganisms Identification using MALDI‐TOF MS

3.3 Factors Impacting the Accuracy of MALDI‐TOF MS Identifications

3.4 Identification of Microorganisms from Positive Cultures

3.5 Identification of Microorganisms Directly from Samples

3.6 Microorganisms Requiring a Specific Processing for MALDI‐TOF MS Identification

3.7 Detection of Antimicrobial Resistance

3.8 Detection of Bacterial Virulence Factors

3.9 Typing and Clustering

3.10 Application of MALDI‐TOF MS in Clinical Virology

3.11 PCR‐Mass Assay

3.12 PCR‐ESI MS

3.13 Impact of MALDI‐TOF MS in Clinical Microbiology and Infectious Disease

3.14 Identification of Protozoan Parasites

3.15 Identification of Ticks and Fleas

3.16 Costs

3.17 Conclusions

References

4 The Challenges of Identifying Mycobacterium to the Species Level using MALDI‐TOF MS

4A.1 Taxonomic Structure of the Genus Mycobacterium

4A.2 Tuberculosis‐Causing Mycobacteria

4A.3 Non‐tuberculosis Mycobacteria

4A.4 MALDI‐TOF MS Mycobacteria Library and Parameters for Identification

4A.5 Methods for Extraction

4A.6 Protein Profiling of Cell Extracts using SELDI‐TOF MS

4A.7 Conclusion

References

Part 4B ASTA’s MicroID System and Its MycoMp Database for Mycobacteria

4B.1 Introduction

4B.2 MycoMp Database for Mycobacterium: The ASTA Mycobacterial Database

4B.3 MicroID Software

4B.4 Database

4B.5 MycoMP Database for Mycobacteria

4B.6 Conclusion

References

5 Transformation of Anaerobic Microbiology since the Arrival of MALDI‐TOF Mass Spectrometry

5.1 Introduction

5.2 Identification in the Clinical Laboratory

5.3 Pre‐analytical Requirements Influence Species Identification of Anaerobic Bacteria

5.4 Recent Database Developments for Anaerobes

5.5 Application of the MALDI‐TOF MS Method for Routine Identification of Anaerobes in the Clinical Practice

5.6 The European Network for the Rapid Identification of Anaerobes (ENRIA) Project

5.7 Subspecies‐Level Typing of Anaerobic Bacteria Based on Differences in Mass Spectra

5.8 Impact of MALDI‐TOF MS on Subspecies Classification of Propionibacterium acnes: Insights into Protein Expression using ESI‐MS‐MS

5.9 Direct Identification of Anaerobic Bacteria from Positive Blood Cultures

References

6 Differentiation of Closely Related Organisms using MALDI‐TOF MS

6.1 Introduction

6.2 Experimental Methods

6.3 Results

6.4 Discussion and Implications

Acknowledgments

References

7 Identification of Species in Mixed Microbial Populations using MALDI‐TOF MS

7.1 Introduction

7.2 A New Algorithm to Identify Mixed Species in a MALDI‐TOF Mass Spectrum

7.3 Toward Direct‐Sample Polymicrobial Identification from Positive Blood Cultures

7.4 In Vitro Experiments

7.5 Discussion and Perspectives

References

8 Microbial DNA Analysis by MALDI‐TOF Mass Spectrometry

8A.1 Introduction

8A.2 The Molecular Detection and Identification of Viruses

8A.3 Viral Quantification

8A.4 The Characterization of Viral Genetic Heterogeneity

8A.5 Viral Transmission Monitoring

8A.6 Additional Nucleic Acid Applications of MALDI‐TOF MS

8A.7 Conclusion

References

Part 8B Mass Spectral Analysis of Proteins of Nonculture and Cultured Viruses

8B.1 Introduction

8B.2 Norovirus Identification using MS

8B.3 Sample Preparation Considerations

8B.4 Experimental Workflow

8B.5 Detection of Intact VP1 using MALDI‐TOF and SELDI‐TOF MS

8B.6 Peptide Mass Fingerprinting

8B.7 Conclusions

8B.8 Bacteriophage Identification using MS

8B.9 Bacteriophages

8B.10 Protein Identification

8B.11 Conclusions

References

9 Impact of MALDI‐TOF MS in Clinical Mycology; Progress and Barriers in Diagnostics

9.1 Introduction

9.2 Evolution in Commercial Methodologies of Sample Preparation

9.3 Effect of In‐House Sample Preparation on Database Reliability

9.4 Conclusion

References

10 Development and Application of MALDI‐TOF for Detection of Resistance Mechanisms

10.1 Attempts to Correlate Signature Mass Ions in MALDI‐TOF MS Profiles with Antibiotic Resistance

10.2 Distribution and Spread of Carbapenems and Mass Spectrometry

10.3 Carbapenem‐Resistant Enterobacteriaceae

10.4 MALDI‐TOF MS Detection Based upon Changes in Antibiotic Structure due to Bacterial Degradation Enzymes

10.5 Optimization of the Carbapenemase MALDI‐TOF MS‐Based Assay to Minimize the Time‐to‐Result

10.6 Detection of Other Bacterial Enzymic Modifications to Antibiotic Structures

10.7 Isotopic Detection using MALDI‐TOF MS

10.8 Multi‐Resistant Pseudomonas aeruginosa

10.9 MALDI Biotyper Antibiotic Susceptibility Test Rapid Assay (MBT‐ASTRA™)

10.10 The Potential Use of Mass Spectrometry for Antibiotic Testing in Yeast

References

11 Discrimination of

Burkholderia

Species,

Brucella

Biovars,

Francisella tularensis

and Other Taxa at the Subspecies Level by MALDI‐TOF Mass Spectrometry

11.1 Introduction

11.2 Principles of MALDI‐TOF MS‐Based Identification of Bacteria

11.3 Generality versus Specificity

11.4 Shigatoxin‐Producing and Enterohemorrhagic Escherichia coli (STEC and EHEC)

11.5 Francisella tularensis

11.6 The Genus Brucella

11.7 The Genus Burkholderia

11.8 Studying Closely Related Organisms by MALDI‐TOF MS

11.9 Conclusion

References

12 MALDI‐TOF‐MS Based on Ribosomal Protein Coding in

S10‐spc‐alpha

Operons for Proteotyping

12.1 Introduction

12.2 S10‐GERMS Method

12.3 Conclusion: Computer‐Aided Proteotyping of Bacteria Based on the S10‐GERMS Method

References

Part II: Tandem MS/MS‐Based Approaches to Microbial Characterization

13 Tandem Mass Spectrometry Analysis as an Approach to Delineate Genetically Related Taxa

Part A

13.1 Introduction

13.2 Methods

13.3 Results

13.4 Candidate Biomarker Discovery: Shotgun Sampling of Enterobacteriaceae Proteomes by GeLC‐MS/MS

13.5 Discussion

Part B

13.6 Highly Pathogenic Biothreat Agents

13.7 Bacillus anthracis

13.8 Summary of Results

13.9 Yersinia pestis

13.10 Method: Strain Panel

13.11 Summary of Results

13.12 Fransicella tularensis

13.13 Method

13.14 Summary of Results

13.15 Clostridium botulinum

13.16 Method

13.17 Summary of Results

13.18 Burkholderia pseudomallei and B. mallei

13.19 Method

13.20 Summary of Results

13.21 Biomarker Detection Sensitivity and Quantification

13.22 Method

13.23 Summary of Results

13.24 Assay Sensitivity in Relation to Bacterial Cell Numbers

13.25 Summary of Results

13.26 Spiked Samples

13.27 Method

13.28 Summary of Results

13.29 Spiked Cells

13.30 Method

13.31 Summary of Results

13.32 B. anthracis Spore Analysis

13.33 Method

13.34 Summary of Results

13.35 Assay Sensitivity in Relation to Bacterial Spore Numbers

13.36 Method

13.37 Summary of Results

13.38 Summary of Results for Biomarker Detection Sensitivity

References

14 Mapping of the Proteogenome of

Clostridium difficile

Isolates of Varying Virulence

14.1 Introduction

14.2 Virulence of Clostridium difficile

14.3 Current Genomic and Proteomic Data

14.4 Comparison of Strains of Varying Virulence

14.5 Genomic Analysis of Clostridium difficile

14.6 Proteomic Analysis of Clostridium difficile

14.7 Mapping the Proteogenome of Clostridium difficile to Phenotypic Profiles

14.8 Antibiotic Resistance

14.9 Conclusion

References

15 Determination of Antimicrobial Resistance using Tandem Mass Spectrometry

15.1 Antibiotic Resistance Mechanisms

15.2 Detection of β‐lactamase Activity

15.3 Other MALDI‐TOF MS Methods

15.4 Liquid Chromatography Coupled with MS

15.5 Proteomics Approaches for Detection of Antibiotic Resistance

15.6 Conclusion

References

16 Proteotyping

16.1 Introduction

16.2 MS and Proteomics

16.3 MALDI TOF MS

16.4 Tandem MS Shotgun Proteomic Analyses

16.5 Top‐Down Proteomics

16.6 Bottom‐Up Proteomics

16.7 Proteotyping

16.8 Matching MS Spectra to Peptides

16.9 Mapping Peptides to Reference Sequences

16.10 Taxonomic Assignment of Protein Sequences

16.11 Challenges Assigning Fragments to Lower Taxonomic Levels

16.12 Proteotyping for Diagnosing Infectious Diseases

16.13 Outlook

16.14 Conclusion

Acknowledgments

References

17 Proteogenomics of

Pseudomonas aeruginosa

in Cystic Fibrosis Infections

17.1 Introduction: Pseudomonas aeruginosa as a Clinically Important Pathogen

17.2 CF and Pathophysiology

17.3 CF Infections

17.4 Biofilm Formation in P. aeruginosa

17.5 Virulence of P. aeruginosa

17.6 Genomics to Study Bacterial Pathogenesis

17.7 Proteomics to Study Bacterial Pathogenesis

17.8 Genomics of P. aeruginosa in CF Infections

17.9 Interclonal Genome Diversity

17.10 Intraclonal Genome Diversity

17.11 Clonal Spread of P. aeruginosa in CF Patients

17.12 Parallel Evolution

17.13 Mutations in Early‐Stage CFP. aeruginosa Isolates

17.14 Mutations in Late‐Stage CF P. aeruginosa Isolates

17.15 Transcriptomics of P. aeruginosa in Chronic CF Infections

17.16 Proteomics of P. aeruginosa in Chronic CF Infections

17.17 Applications of Proteomics to P. aeruginosa Characterization

17.18 Comparative Proteomic Investigation of Bis‐(3′‐5′)‐Cyclic‐Dimeric‐GMP (C‐Di‐GMP) Regulation in P. aeruginosa

17.19 Comparative Proteomics of Mucoid and Non‐Mucoid P. aeruginosa Strains

17.20 Proteogenomics Reveal Shifting in Iron Uptake of CFP. aeruginosa

17.21 Conclusion and Future Perspectives

References

18 Top‐Down Proteomics in the Study of Microbial Pathogenicity

18.1 Introduction

18.2 Top‐Down Analysis of Modified Bacterial Proteins in Targeted Mode

18.3 Top‐Down Analysis of Bacterial Proteins in Discovery Mode

18.4 Top‐Down Proteomics: The Next Step in Clinical Microbiology?

References

19 Tandem Mass Spectrometry in Resolving Complex Gut Microbiota Functions

19.1 Introduction

19.2 MS in Microbiology

19.3 Intestinal Metaproteomics Addressing All Proteins

19.4 LC‐MSMS Analysis

19.5 Data Analysis

19.6 Data Output and Interpretation

19.7 Development of Surface Metaproteomics for Intestinal Microbiota

19.8 Conclusions

References

20 Proteogenomics of Non‐model Microorganisms

20.1 Introduction

20.2 The “Proteogenomics” Concept

20.3 Applications to Non‐model Organisms: From Bacteria to Parasites

20.4 Embracing Complexity with Metaproteogenomics

References

21A Analysis of MALDI‐TOF MS Spectra using the BioNumerics Software

21A.1 Introduction

21A.2 Typing with MALDI‐TOF MS

21A.3 Preprocessing of Raw MALDI‐TOF MS Data

21A.4 Downsampling

21A.5 Baseline Subtraction

21A.6 Curve Smoothing

21A.7 Peak Detection

21A.8 Biological and Technical Replicates

21A.9 Averaging of Replicates

21A.10 Spectrum Analysis

21A.11 Hierarchical Clustering

21A.12 Alternatives to Cluster Analysis

21A.13 Classifying Algorithms

21A.14 Conclusion

References

21B Subtyping of

Staphylococcus

spp. Based upon MALDI‐TOF MS Data Analysis

21B.1 Introduction

21B.2 Sample Collection

21B.3 MALDI‐TOF Mass Spectrometry

21B.4 Cluster Analysis of Environmental Staphylococci

21B.5 Antibiotic Susceptibility Test

21B.6 Cluster Analysis of Staphylococcus spp. Recovered from Different Sites

21B.7 Correlation of Staphylococci Recovered from Different Sites

21B.8 Cluster Analysis of S. epidermidis Isolated from Different Sites

21B.9 Cluster Analysis of S. aureus Isolated from Different Sites

21B.10 Cluster Analysis of Staphylococcus spp. Combined with Antibiotic Susceptibility

21B.11 Antibiotic Resistance Patterns of Closely Related S. epidermidis

21B.12 Antibiotic Resistance Patterns of Closely Related S. aureus

21B.13 Variations of Antibiotic Susceptibility of Closely Related S. epidermidis

21B.14 Percentage of Multiple‐Resistant Staphylococci Recovered from Each Site

21B.15 Conclusion

References

21C Elucidating the Intra‐Species Proteotypes of

Pseudomonas aeruginosa

from Cystic Fibrosis

21C.1 The Emergence of Pseudomonas aeruginosa as Key Component of the Cystic Fibrosis Lung Flora

21C.2 Diversity and Rational for Proteotyping

21C.3 Selecting Representative Strains for Profiling

21C.4 Selection of Strains against a Background of Their Variable Number Tandem Repeat (VNTR) Designation

21C.5 Potential to Type P. aeruginosa using MALDI‐TOF MS

21C.6 Data Processing: Analyzing Data using BioNumerics 7

21C.7 Discussion and Data Interpretation

21C.8 Going Forward – Reproducibility the Salient Determinant

References

Index

End User License Agreement

List of Tables

Chapter 01

Table 1.1 International conferences organized by the Molecular Identification Services Unit (MISU) and Applied and Functional Genomics Unit (AFGU) to showcase work achieved and future directions in proteomics using MALDI‐TOF MS and tandem MS/MS and genomics. (MISU and AFGU were amalgamated in September 2009 in the new Department of Bioanalysis and Horizon Technologies.)

Chapter 02

Table 2.1 Different requirements of MALDI‐TOF MS measurements and QC for routine identification and database establishment.

Chapter 03

Table 3.1 Example of interpreted results as currently reported in our laboratory.

Table 3.2 MALDI‐TOF MS applications in clinical microbiology and references.

Table 3.3 Added value of MALDI‐TOF MS in clinical microbiology.

Chapter 04a

Table 4A.1 List of mycobacterial strains used blindly for MALDI‐TOF MS identification.

Chapter 04b

Table 4B.1 The type strains of

Mycobacteriium

spp. from KCTC (Korean Collection of Type Cultures) used for the MycoMP database.

Table 4B.2 103 Clinical strains of Mycobacteria belonging to 25

Mycobacteriium

species.

Chapter 05

Table 5.1 The influence of exposure to oxygen on the quality of the MALDI‐TOF MS spectrum.

Table 5.2 The range of log scores and identification results of spotting the same strain ten times by two different examiners.

Table 5.3 Identification of anaerobic bacteria by MALDI‐TOF MS in clinical microbiology laboratories (evaluation according to the manufacturer’s instructions).

Chapter 06

Table 6.1 Semiautomated model peak frequency profiles across four organism classes.

Table 6.2 Biomarker peaks (

m/z

) identified in the automated MALDI‐TOF MS approaches.

Table 6.3 Accuracy of the hybrid MALDI‐TOF MS assay, serogrouping, and automated biochemicals (Phoenix) relative to the reference identification for test isolates.

Table 6.4 Discrepant results between reference identification and hybrid MALDI‐TOF MS results.

Chapter 07

Table 7.1 List of species considered for the proof of concept related to direct identification from microbial cells extracted from positive blood culture bottles.

Chapter 08b

Table 8B.1 A selection of viruses that have been investigated using proteomic‐based approaches.

Table 8B.2 Observed and theoretical

m/z

of peptides detected using MALDI‐TOF MS analysis of VLP tryptic digest together with error in ppm. The sequence of peptides highlighted in red was confirmed using tandem MS.

Table 8B.3 Observed and theoretical

m/z

of peptides detected using LC‐MS/MS analysis of VP1 originating from VLP‐spiked feces together with error in ppm. The sequence of peptides highlighted in red was confirmed using tandem MS.

Table 8B.4 Proteins identified using LC‐IT‐TOF in MS/MS mode following in‐solution tryptic digestion of bacteriophages protein lysates.

Chapter 12

Table 12.1 Comparison of the similarities of

Pseudomonas putida

at strain level.

Table 12.2 Ribosomal protein profiling table of

P. syringae

strains.

Table 12.3 Theoretical masses of nine selected ribosomal subunit proteins of

Sphingomonacecae

.

Table 12.4 Binary peak matching profile of

B. subtilis

strains.

Table 12.5 Peak pattern of selected biomarkers for the

L. casei

group.

Table 12.6 Peak pattern of selected biomarkers for

E. coli

discrimination.

Chapter 13

Table 13.1 The strains used in this example are listed below, where * = type strain, D = strains used to construct the database and T = strains used to test blindly whether they can be identified using the workflow presented in this study.

Table 13.2 Comparison of identifications generated by 16S rRNA (RDP) and MALDI‐TOF (Biotyper). No ID = unable to determine genus or species. Bold highlighted = incorrectly identified species when compared to the NCTC designation.

Table 13.3 Comparison of identification methods: 16S rRNA, Biotyper and the DB‐FP database. Poorly resolved identifications are in bold.

Table 13.4 Clustering of protein identifications identified by each sample cohort, where outbreak = OB strain, EHEC and EAEC represent OB strain‐related pathovars. Numbers = unique peptide identifications per parent protein (white background = 0, pink = 1 and red = ≥ 2 peptides identified per parent protein).

Table 13.5 Primer sequences used for PCR amplification and DNA sequencing of each target biomarker region.

Table 13.6 Functional categorization of

B. anthracis

–specific peptide markers (* = peptide derived from

B. anthracis

unique proteins,

§

 = plasmid origin,

†

 = no suitable primer region found and

‡

 = possessed a silent mutation).

Table 13.7 Summary of

Y. pestis

–specific peptides. In silico genetic validation of the biomarkers found all markers to be genetically stable.

Table 13.8 Summary of

F. tularensis

–specific peptides and subspecies‐specific peptides. In silico genetic validation of the biomarkers found that all markers were genetically stable.

Table 13.9 Primer sequences used for group‐specific PCR amplification and DNA sequencing of each target biomarker region.

Table 13.10 A summary of

C. botulinum

toxin‐specific and group‐specific markers that were genetically validated by PCR and direct sequencing (* = no suitable primer region found,

§

 = experimentally validated stable marker,

†

 = genetically stable by in silico analysis,

◊

 = silent mutations present,

‡

 = PCR amplification/sequencing not achieved for all relevant

C. botulinum

strains).

Table 13.11 Primer sequences used for PCR amplification and DNA sequencing of each target biomarker region.

Table 13.12 Summary of

B. mallei

– and

B. pseudomallei–

specific peptides subjected to PCR and direct sequencing (* = no suitable primer region found,

§

 = genetically stable marker,

†

 = genetically stable by in silico analysis,

‡

 = amplification not achieved for all

B. pseudomallei

strains). Markers were not found to contain any silent mutations by in silico analysis.

Table 13.13 Selected peptides, their corresponding stable isotope‐labelled standard peptides and retention time.

Table 13.14 Abundance of target peptides in selected

B. anthracis

strains. Each value is the mean value ± S.E. of three technical replicates.

Table 13.15 Selected peptides, their corresponding stable isotope‐labelled standard peptides, retention time and abundance of target peptides in selected

Clostridium botulinum

strains.

Table 13.16 Summary of

B. anthracis

–specific peptides.

Table 13.17 Summary of

B. anthracis

–specific proteins and peptides that were detected from 10

5

B. anthracis

spores.

Table 13.18 Summary of results for biomarker detection sensitivity and quantification.

Chapter 14

Table 14.1 Roche assembly metrics. A comparable number of open reading frames (ORFs) were identified in strains 027 SM (3896) and Tra 5/5 (3840), similar to the number identified in the 630 strain. However, a larger number of ORFs (4061) were identified in strain B‐1. Extra‐chromosomal data were generated for three plasmids in strain B‐1, and one plasmid in strain Tra 5/5.

Table 14.2 PacBio sequencing metrics.

Table 14.3 To determine whether, based on MIC profiles like those above, an isolate is resistant or sensitive to a particular antibiotic, external bodies such as the British Society for Antimicrobial Chemotherapy (BSAC) offer guidance notes on MIC cut‐offs, referred to as ‘MIC break points’. For example, the MIC breakpoints for the commonly prescribed (for

C. difficile

infections) antibiotics, metronidazole and vancomycin, are ‘sensitive ≤2 mg/l and resistant >2 mg/l’. In the context of the three described strains, A027, B‐1 and Tra5/5, all three are sensitive and thus treatable using either metronidazole or vancomycin.

Chapter 15

Table 15.1 Timeline of the discovery, introduction and first observed resistance of antibiotics.

Table 15.2 Comparison of MALDI‐TOF MS and LC‐MS.

Chapter 16

Table 16.1 Overview of different mass analyzers and their analytical performance.

Chapter 19

Table 19.1 Items determining decision making when selecting either MALDI‐ or tandem‐MS‐based technology.

Table 19.2 Overview of fecal metaproteomic studies in humans.

Table 19.3 The 10 bacterial protein clusters detected with highest Mascot scores in biotin‐enriched surface preparations of fecal bacteria. The proteins with over 80% homology with each other were clustered together. Mascot score shown is the maximal score of the proteins within a cluster. Cluster size indicates number of proteins clustered together.

Chapter 21-1

Table 21A.1 Description of the two workflows used for comparative data analysis of the different preprocessing algorithms.

Table 21A.2 Results (P for precision, R for recall) of the comparison of five different classifying algorithms on four different datasets.

List of Illustrations

Chapter 01

Figure 1.1 Examples of the distinctive MALDI‐TOF‐MS profiles of intact cells of

Porphyromonas

sp. containing both genus‐specific (e.g. 618 and 844 Da mass ions) and also a significant number of species‐specific mass ions (examples indicated by arrows). Members of the genus

Porphyromoas

now comprise 18 species, in addition to several others that have not yet been validated (Bergey’s manual, 2011). However, with the exception of DNA/DNA reassociation, they could not be reliably delineated at the time. The three representative MALDI‐TOF‐MS spectra shown and those reported earlier (Shah

et al.

, 2002) revealed that each species could be unambiguously distinguished. It was this poorly characterized group of anaerobes that became one of the compelling forces for the development of this technique for microbial identification. Early meetings to demonstrate an appreciation of the technology were mostly presented at meetings on anaerobic taxa and helped rejuvenate interest in this area of microbiology again (see Chapter 5 and Figure 1.4).

Figure 1.2 The first bench‐top linear MALDI‐TOF mass spectrometer, the Kratos Kompact Alpha (Kratos Analytical, Manchester, UK), that inspired the development of the technology for clinical microbiology. The target plate, with a capacity for 20 samples, is shown and was used for generating the spectra shown in Figure 1.1. The instrument revealed the potential of the technology, but it was manual and unsuitable for a microbial diagnostic laboratory.

Figure 1.3 The conference leaflet of the first meeting held on 27 October 1998 to explore the potential application of MALDI‐TOF mass spectrometry for microbial identification. These meeting were held annually to showcase developments in the field. However, it still took a decade of research and development for this technology to gain widespread acceptance in the clinical microbiology laboratory.

Figure 1.4 Changes in the MALDI‐TOF‐MS profile of the same strain of

Porphyromonas catoniae

(NCTC 12856) grown of Fastidious Anaerobic Agar (FAA) and Columbia Blood Agar (CBA). Many of the significant mass ions, e.g. 542, 580, 618, 689, 784 and 845 Da, are retained. However, significant mass ions such as 935 Da are present only in cells grown on CBA.

Figure 1.5 The first dedicated linear MALDI‐TOF mass spectrometer for microbial identification used for creating the first MALDI‐TOF MS database in 2004 (Keys

et al.

, 2004). The instrument, manufactured by Micromass (Manchester, UK), remedied many of the shortcomings of the Kratos Konpact Alpha and had the capacity to analyze 96 samples automatically.

Figure 1.6 Early meetings on MALDI‐TOF MS were focused on poorly defined anaerobic species to emphasize, even at that time, the resolution of the method. Many of these species are non‐fermentative so that methods involving the use of API or various biochemical tests, which were the primary methods then, were negative. New species were proposed mainly on the basis on DNA‐DNA reassociation, which could not be applied in a clinical laboratory. MALDI‐TOF MS when introduced contributed significantly to a resurgence of interest in anaerobic microbiology because of its capacity to resolve such complex taxonomic problems. (A) As early as the year 2000, a specific symposium was held to demonstrate the high resolving power of MALDI‐TOF MS in delineating poorly defined anaerobic species in the United Kingdom and (B) in 2001, in the United States.

Figure 1.7 The first promotional service flyer for the Molecular Identification Services Unit (MISU, PHLS), which was printed and advertised from 2000. It signalled a strong declaration of MISU’s vision to elevate MALDI‐TOF MS as its principal method of microbial identification by placing a M@LDI‐TOF mass spectrometer (Micromass, Manchester, UK) in the central position on the flyer.

Figure 1.8 Identification of clinical isolates received by MISU using 16S rRNA and MALDI‐TOF MS (Micromass, UK) over a 10‐year period. MISU receives atypical, rarely isolated and emerging pathogens, and for such unusual isolates, the results shows that MALDI‐TOF MS was significantly more useful in identifying these isolates to the species level.

Figure 1.9 Some of the most popular types of ProteinChip arrays (H50, Q10, CM10 and NP20) used for analysis of microbial cells extracts and an overview of the process to obtain a mass spectral profile. From left to right, the sample is added to the ProteinChip array, the wells washed and air‐dried for a few minutes, the matrix (Sinapinic acid) added, followed by mass spectral analysis to yield the spectra shown.

Figure 1.10 Proposed mechanism for the degradation of DNA in a MALDI‐TOF mass spectrometer. This was taken from Franz Hillenkamp’s notebook during his visit to MISU and AFGU in 2008. See text for details.

Figure 1.11 Identification of

Clostridium difficile.

Initially this posed a major problem for accurate and reproducible identification. The dendrogram shows some100 clinical isolates of

C. difficile

clustering in a single phenon and distant from other species such as

Propionibacterium acnes

and

Staphylococcus warneri

, with which it initially formed a common cluster.

Figure 1.12 Scanning electron micrographs of

Clostridium difficile

cells mixed with the matrix solution 2,5‐ dihydroxy benzoic acid in acetonitrile: ethanol: water (1:1:1) with 0.3% TFA. The electron micrographs show the clumps of cells which give a MALDI‐TOF MS spectrum (indicated by positive) if the laser strikes, whereas in areas where there are no cells, no spectra are obtained (negative).

Figure 1.13 Bottom‐up workflow used to deduce strain‐specific peptides and virulence determinants of

E. coli

O104:H4 during the outbreak of 2011. The classical 1‐D SDS‐PAGE followed by in‐gel trypsin digestion or directly loaded onto LC‐MS/MS were also used. Three different Orbitrap mass spectrometers, with varying resolutions, were used to obtain comprehensive profiles of the proteome of strains.

Figure 1.14 Phasing in of MS‐based‐methods in MISU‐AFGU into clinical microbiology from 1998. These methods were used to characterize clinical isolates by our laboratories. The time lines and providers of the technology are shown in the figures. In 2010 bioMérieux acquired AnagnosTec, while Sequenom’s interest shifted away from microbiology to solely human applications.

Figure 1.15 A comparative overview of bottom‐up versus top‐down proteomics. Although the former has been used extensively for comprehensive analysis of the proteome of several species, it is currently too cumbersome for a clinical laboratory. It seems likely that top‐down based approaches will supersede MALDI‐TOF MS as the next‐generation approach to microbial identification and typing.

Figure 1.16 Distribution of cellular proteins of a cell extract to assess the efficacy of an extraction method. The figure shows the protein distribution of a standard

E. coli

strain using mechanical breakage and centrifugation steps. Methods such as sonication released mainly ribosomal proteins and yields profiles similar to MALDI‐TOF MS but more complex. A method that captures a broad cross‐section of protein is, in our view, far more beneficial for a future database.

Chapter 02

Figure 2.1 Example of spectra acquired with different laser energies. Too low laser energy leads to few peaks with low intensity, and too high energy causes a lift of baseline and broadening of peaks. Broadening of peaks can lead to fusion of closely located signals.

Figure 2.2 Dendrogram of a novel MSP (strain VA2219, depicted in red) with old database references; wrong pre‐identification has been detected by cluster analysis. The strain could be confirmed as

Aggregatibacter aphrophilus

, subsequently.

Chapter 03

Figure 3.1

Principle of MALDI‐TOF MS identification of microorganisms.

The sample is first deposited on a metal plate and embedded in the matrix that crystallizes the analytes; it is then bombarded by brief laser pulses that achieve the ionization (MALDI) by proton transfer from the matrix, which results in positively charged analytes. The desorbed ions are then accelerated by an electrostatic field and directed in the flight tube, in which they are separated according to their time of flight (TOF) in the flight tube, in which a high vacuum is generated by a pump. Ions are detected at the exit of the flight tube, and a software generates a mass spectrum. The identification is achieved by comparison of the mass spectrum with a database of reference mass spectra

Figure 3.2 Application of MALDI‐TOF MS in clinical microbiology.

Chapter 04a

Figure 4A.1 Phylogenetic structure of the genus Mycobacterium. The neighbour‐joining tree is based on 16S sequences from 17 smooth mycobacterial and MTBC strains. The blue triangle represents the MTBC strains, which differ by up to one nucleotide. Bootstrap support higher than 90% shown on nodes. Scale bar is pairwise distances after Jukes‐Cantor correction. Image reproduced from Gutierrez

et al.

(2005), under the Creative Commons Attribution License (CCAL), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Figure 4A.2 Electron microscopy images of

Mycobacterium fortuitum

cells: (A) Native cells (cell suspension in water diluted 1:1 in a fixing solution), (B) cells after inactivation (30 min at 95°C in a water bath), (C) pellet of inactivated cells after mechanical lysis with 0.5 mm silica beads, (D) cells after lipid extraction with dichloromethane (before treatment with ethanol and bead beating); loss of polarity and clumping of cells were evident.

Figure 4A.3 Influence of growth and methods on identification scores.1: (Top) MGIT broth, 0.1 mm silica beads, elution in 100% acetonitrile and 70% formic acid. Score: 1.881 2: (Middle) LJ culture, 0.1 mm silica beads, elution in 100% acetonitrile and 70% formic acid. Score: 1.7823: (Bottom) As in A above, LJ culture, 0.5 mm beads. Score: 2.066

Figure 4A.4 Partial mass spectrum using SELDI‐TOF MS of

Mycobacterium

species. From top to bottom;

M. abscessus, M. agri. M. fortuitum, M. intracellulare, M. kansasii

and

M. smegmatis.

SELDI‐TOF MS has the capacity to generate mass ions in excess of 100,000 kDA (see Shah

et al.

, 2005) and clearly has potential application for identification of

Mycobacterium

species.

Chapter 04b

Figure 4B.1 The Micro ID system comprises the ASTA Tinkerbell LT MALDI‐TOF spectrometer, Micro ID software and database, and disposable kit with disposable MALDI plates, matrix, standards solution.

Figure 4B.2 Typical spectra obtained by MALDI‐TOF‐MS (Tinkerbell LT, Suwon, Korea) of

Mycobacterium

species: (A)

Mycobacterium gordonae

, (B)

Mycobacterium fortuitum

, (C)

Mycobacterium saskachewanense

and (D)

Mycobacterium phlei.

Figure 4B.3 MALDI spectra representing the

Mycobacterum tuberculosis

complex: (A)

Mycobacterium tuberculosis

, (B), (C), (D)

M. bovis

BCG.

Figure 4B.4 Structure of MicroID database. Each database (DB) is dedicated to a specific application.

Figure 4B.5 The workflow for developing a mycobacterial MALDI‐TOF MS database.

Figure 4B.6 Stages in the preparation of a mycobacterial sample for MALDI‐TOF MS analysis.

Figure 4B.7 Cluster analysis of selected mass spectra of

Mycobacterium

spp. from the MycoMP database using Perseus (Computational Systems Biochemistry, Germany). The inter‐relatedness of the

Mycobacteriium

species is illustrated and

M. tuberculosis

is distinguished from other mycobacteria.

Chapter 05

Figure 5.1 (Left): Partial MALDI‐TOF MS profiles of anaerobic culture cell extracts of

P. acnes

types and other species. Types I and II share more common mass ions compared to type III. (Right): SELDI profiles of aerobic and anaerobic cultured cell extracts

P. acnes

types covering a mass range between 5 kDa to 30 kDa,

Figure 5.2 Cell morphologies of three types revealed by electron microscopy. A: Type IB (strain K115), B: Type II (strain B23), C: Type III (strain B12) showing an elongated cell structure (0.4–0.7 µm and up to 15 µm long).

Figure 5.3 1D SDS‐PAGE gel analysis of type IB (strain K115). The anaerobic and microaerophilic culture lysates showed dense bands at 12–15 kDa (1 and 3) compared with same section of aerobic lysate band (2). These were excised, trypsin‐digested and analyzed using ESI‐MS/MS (see text for details).

Chapter 06

Figure 6.1 Representative MALDI Biotyper analysis of a

S. sonnei

isolate. The database lacks

Shigella

spp. reference spectra, and highly reliable species‐level matches occur for

E. coli

.

Figure 6.2 Pearson’s correlation coefficients for isolates tested by the four‐class semiautomated model (Table 6.1). Three‐dimensional plot showing the degree of correlation between each isolate and the four‐peak frequency profiles (Table 6.1):

S. flexneri

(Sf),

S. sonnei

(Ss), typical

E. coli

(Ec), and inactive

E. coli

(iEc). A score of +1 indicates a direct correlation and −1 indicates a perfect inverse correlation.

Figure 6.3 An example of peaks chosen by the ClinProTools automated model generation algorithm. Spectra are shown in a gel‐like view with

E. coli

spectra below and

Shigella

spp. spectra above the dotted line. Peak 8349 was chosen by the algorithm for the genus‐level model, 8324 for the species‐level model, and 8366 was considered nondiscriminatory.

Figure 6.4 Peaks likely corresponding to amino acid changes in the acid resistance chaperone HdeA.

Chapter 07

Figure 7.1

Artificial mixtures prepared in vitro.

Artificial mixtures were prepared in vitro by diluting and mixing two calibrated monobacterial suspensions to obtain bimicrobial suspensions at the following relative concentrations: 1:0, 10:1, 5:1, 2:1, 1:1, 1:2, 1:5, 1:10, and 0:1. (A)

Visualization of mixed mass spectra.

Five smoothed mass spectra obtained at concentrations 1:0, 2:1, 1:1, 1:2, and 0:1 are shown on a 2–12 kDa grid. The top (resp. bottom) panel represents the pure spectra of the first (resp. second) species of the mixture, and the panels in between represent the spectra obtained when the proportion of the second species increases. We note that peaks specific for the second species gradually appear and increase with its relative proportion, whereas the peaks specific of the first species gradually decrease and disappear. (B)

Mixed versus reference mass spectra

. A principal component analysis (PCA) was carried out from the spectra of the reference database corresponding to the two species involved in these mixtures, shown as red and blue empty circles. The spectra obtained from the in vitro mixtures were then projected in the PCA space and shown as filled circles, with the color turning from red to blue as the relative proportion of the second species increases. We note a remarkably smooth transition from the first to the second species as their relative concentration varies. This PCA analysis was carried out from peak‐list representation of the spectra. These artificial mixed spectra correspond to the mixture E involved in Mahé

et al.

, 2014, and are available online at the UCI Machine Learning Repository (https://archive.ics.uci.edu/ml/datasets/MicroMass).

Figure 7.2

The mixed identification algorithm.

Shown on the right‐hand side (orange) are the operations done off‐line to prepare the species‐specific prototypes. First, a prototype must be built for each reference species considered (shown here in red, blue, and green for species A, B, and C). These prototypes are derived from the reference spectra database embedded in the commercial VITEK®MS system. Then, although not strictly mandatory, the prototypes should be adjusted in order to limit the risk of obtaining erroneous decompositions in subsequent analyses. Shown on the left‐hand side (blue) are the operations carried out to predict the composition of a sample on the basis of these reference prototypes. First, a spectrum is acquired and preprocessed by the standard algorithm used for routine (culture‐based) identification, in order to extract its prominent peaks. A list of candidate (poly)microbial compositions is then obtained by means of a nonnegative linear regression framework involving the Lasso penalty and the LARS algorithm. Finally, the most plausible decomposition, achieving a trade‐off between the accuracy of the approximation of the spectrum as a combination of prototypes and the number of species involved, is selected by the Bayesian information criterion (BIC).

Figure 7.3

In silico results – simulated mixtures.

(A)

Overall performance

in terms of correct (green), partial (orange), and erroneous (red) identification computed globally (left), and on mixed and pure spectra only (middle and right, respectively). The proportions of erroneous identification are split into same‐genus misidentifications that involve prediction of species that belong to the same genera as the ones actually present in the samples (light red, second figure shown on top of the bar) and those involving species of other genera (dark red). (B)

Mixed spectra:

Species‐level performance. Performance obtained among mixed spectra involving each of the species of the reference dataset. (C)

Same‐genus misidentifications

: Proportion of genera involved. The great majority of same‐genus errors involve confusing streptococci and to a lesser extent staphylococci. (D) Streptococci

confusion

. The great majority of same‐genus errors involving

Streptococci

amount to confusing

Streptococcus oralis

and

S. pneumoniae

.

Figure 7.4

In vitro results – spiked monomicrobial blood cultures.

(A)

Spectrum detection and identification performance.

Overall performance of the algorithm in detecting and identifying spiked monomicrobial blood cultured processed according to the reference (left) and direct‐sample (right) protocols. Each bar represents the proportion of the spectral dataset for which a given number of species (here, 1 or 2) was predicted. The green and gray fractions of the bars represent the proportions of spectra correctly and misidentified, respectively. (B)

Overview of misidentifications.

Details about the misidentifications obtained using the reference protocol (left) and the direct‐sample protocol (right). In both cases, the top row gives the actual relative concentration of the species in the mixture, and the bottom row gives the relative concentrations inferred by the algorithm, obtained by normalizing the vector

γ

of coefficients obtained by the algorithm, as described in the text.

Figure 7.5

In vitro results – mixed positive blood cultures.

From top left to bottom right: mixtures A, B, C, and D. In each case, the top row gives the actual relative concentration of the species in the mixture (100%–0%, 90%–10%, 50%–50%, 10%–90%, or 0%–100%), and the bottom row gives the relative concentrations inferred by the algorithm, obtained by normalizing the vector

γ

of coefficients obtained by the algorithm, as described in the text.

Chapter 08a

Figure 8A.1 Workflow of PCR/primer extension MALDI‐TOF MS assays. PCR and primer extension with dNTP/ddNTP stop mixes (left); PCR and primer extension with mass‐modified deoxyribonucleotides (right).

Figure 8A.2 Schematic workflow of PCR, T7‐, and SP6‐ directed in vitro transcription and base‐specific cleavage for comparative sequencing.

Chapter 08b

Figure 8B.1 Workflow for sample preparation and analysis of VP1 protein from norovirus.

Figure 8B.2 Mass spectra showing VP1 detected using (A) MALDI‐TOF MS and (B) SELDI‐TOF MS. In the MALDI‐TOF spectrum, only the single‐charged monomer was detected. In the SELDI‐TOF spectrum, besides the peak corresponding to the single‐charged monomer (59,759

m/z

Da), peaks corresponding to the double‐charged monomer (29,932

m/z

Da) and single‐charged dimer (120,458

m/z

Da) of VP1 were also detected.

Figure 8B.3 Gel electrophoresis showing (left) separation of VP1 protein and proteins captured from the clean VLP suspension and (right) those in fecal samples spiked with VLPs.

Figure 8B.4 PMF spectra of bacteriophage lambda major capsid protein digests showing the differences between (A) in‐solution and (B) in‐gel digestion. The spectra were recorded using MALDI‐TOF MS in positive ion reflectron mode.

Chapter 09

Figure 9.1 Spectra obtained for

Aspergillus udagawae

DMic 062879 (A and B) and

Trichosporon asahii

CBS 2479 (C and D) by the MALDI Biotyper® system. Spectra A and C were obtained by the MALDI Biotyper standard protocol, and spectra B and D were obtained by the NIH‐National Institutes of Health protocol.

Figure 9.2 Dendrograms obtained for species of

Trichosporon

by (A) molecular biology through the IGS1 region sequencing; by MALDI Biotyper® using (B) the manufacturer’s protocol (ethanol‐formic acid‐acetonitrile); and (C) using protein extraction by treatment of the sample with trypsin (Credits are due to Elaine Francisco (UNIFESP, Brazil] and Cledir Santos (Universidad de La Frontera, Chile].)

Figure 9.3 Analysis of the mass spectral profiles obtained by SARAMIS™ software package for (A)

T. rubrum

strains and (B) its comparison with molecular ITS sequence.

Figure 9.4 Dendrogram obtained for DMic strains with data generated by the MALDI Biotyper® from applying the manufacture’s protocol with ethanol‐formic acid‐acetonitrile in blue and the NIH protocol with absolute ethanol and zirconia‐silica beads in yellow.

Chapter 10

Figure 10.1 Mass spectrum of USA 100 MRSA strain. PSMmec peak (2415 m/z) as MRSA marker and the delta‐toxin (3007 m/z) as surrogate marker for agr system (Josten

et al.

, 2014).

Figure 10.2 Representative spectra for p019 detection. Peak intensity is displayed in arbitrary units (10

4

A.U.). Presence of the ~11,109 Da peak in

K. pneumoniae

(A) and

E. coli

(B) p019‐containing isolates (Youn

et al.

, 2016).

Figure 10.3 (A) Basic principle of the MS ß‐lactam hydrolysis assay. The ß‐lactam ring is hydrolyzed (+18 m/z), and the unstable reaction product is decarboxylated (−44 m/z). (B) Intact ertapenem (2nd panel) with 476 Da and its two sodium salts (499/522 Da), and the degraded product (4th panel) with 450 Da (−26 m/z) (Burckhardt and Zimmermann, 2011).

Figure 10.4 Workflow for conventional infection control screening for CRE (left panel) and with automated streaking and ID/AST determination by mass spectrometry (right panel).

Figure 10.5 Workflow of the ß‐lactam hydrolysis assay as proposed in Sparbier

et al.

, 2012.

Figure 10.6 Basic principles of the MS‐Resist™ assay using broth with heavy non‐radioactive lysine as metabolic marker (upper panel). An oxacillin‐susceptible

S. aureus

strain (left) and an MRSA strain (right) mass spectra (pseudo‐gel view) after 3 h incubation (Sparbier

et al.

, 2013).

Figure 10.7 Workflow of the MBT STAR‐ASTRA™ assay (upper panel) and examples with a meropenem‐susceptible

K. pneumoniae

strain (left lower panel) and a resistant KPC strain (right lower panel; pseudo‐gel view) as described in (Sparbier

et al.

, 2016).

Chapter 12

Figure 12.1 Advantage of using ribosomal proteins in comparison with 16S rRNA.

Figure 12.2 Construction procedures of the working database for MALDI‐TOF MS analysis based on the

S10

‐GERMS method.

Figure 12.3 Calculation procedure of theoretical masses of ribosomal protein processed by

N

‐terminal methionine loss.

Figure 12.4 Representative MALDI‐TOF mass spectra of ribosomal proteins of type strains of genus

Pseudomonas

encoded in

S10‐spc‐alpha

operon: (A)

P. putida

NBRC 14164

T

, (B)

P. alcaligenes

NBRC 14159

T

, (C)

P.chlororaphis

NBRC 3904

T

, (D)

P. syringae

NBRC 14083.

Figure 12.5 MALDI‐TOF mass spectra of important biomarkers, S14, L24, and S13 for discrimination of

P. Putida

at the strain level.

Figure 12.6 Phylogenetic analysis of 12 strains of

P. putida

: (A) phylogenetic analysis of

gyrB

sequences, (B) phylogenetic analysis based on the

S10

‐GERMS method.

Figure 12.7 Phylogenetic trees of APEO

n

‐degrading bacteria and type strain of genus

Sphingomonas

based on (A) 16S rRNA gene sequences and on (B) the

S10

‐GERMS method, and (C) MALDI mass spectra of ribosomal protein S14 of

Sphingopyxis terrae

: NBRC 15098

T

, APEO

n

‐degrading bacterium strain BSN20, NBRC 15593, NBRC 15598, and NBRC 15599, respectively.

Figure 12.8 MALDI mass spectra of

B. subtilis

subsp.

subtilis

NBRC 13719

T

using acid extraction: (A) TFA1.0%, (B) TFA2.5, (C) FA30%.

Figure 12.9 Phylogenetic trees of type strains of genus

Bacillus

strains: phylogenetic tree based on amino acid sequences of eight selected ribosomal proteins in (A)

S10‐spc

operon, (B) phylogenetic tree based on 16S rRNA gene sequence.

Figure 12.10 Phylogenetic trees of

B. subtilis

strains based on ribosomal protein profile matching using ribosomal protein of

B. subtilis

subsp.

subtilis

NBRC 13719

T

as a reference strain: (A) phylogenetic tree based on eight ribosomal protein biomarkers, (B) phylogenetic tree based on 20 ribosomal protein biomarkers.

Figure 12.11 The UPGMA cluster analysis of the

L. casei

group based on ribosomal protein biomarkers. Underlined strains are genome‐sequenced. Clusters were divided at a similarity of 80%. The result of ribotyping reported by Ryu

et al.

[92]is used as a reference.

Figure 12.12 Typical MALDI mass spectra of four biomarker proteins in

E. coli

: HdeB (

m/z

9066.2 [M + H]

+

), S15 (

m/z

10138.6/10166.6 [M + H]

+

), L25 (

m/z

10676.4/10694.4 [M + H]

+

), and H‐NS (

m/z

15409.4/15425.4 [M + H]

+

) for strains O 157, O26, O111, and other

E. coli

(non‐O157).

Figure 12.13 Cluster analysis for

E. coli

strains with selected biomarkers. Phylogenetic tree is made based on Table 12.6. A to P indicate the

E. coli

groups classified in Table 12.6.

Figure 12.14 Strategy for MALDI‐TOF MS proteotyping between

E. coli

strains O157, O26, O111, and the others using four biomarkers.

Figure 12.15 Discriminating mixtures of two different types of

E. coli

strains. Strains of K12 and O157 were used as non‐O157 representative and O157, respectively. The

m/z

values of the biomarkers are S15 (10138) and L25 (10694) in

E. coli

K 12, and S15 (10166) and L25 (10676) in

E. coli

O157 strain, respectively. The mixed ratio of K12:O157 is referred to at the right. This figure was modified based on Reference 105.

Figure 12.16 The main differences between the

S10

‐GERMS method and other conventional methods.

Figure 12.17 Concept of the automatic proteotyping system combining conventional MALDI‐TOF MS fingerprinting with the

S10

‐GERMS method.

Chapter 13

Figure 13.1 Dendrogram describing MALDI spectra similarity between isolates in the Enterobacteriaceae database.

Figure 13.2 The distribution of COG functional categories, for all parent proteins, in the Enterobacteriaceae database, where: J = translation, ribosomal structure and biogenesis, A = RNA processing and modification, K = transcription, L = replication, recombination and repair, D = cell cycle control, cell division and chromosome partitioning, V = defence mechanisms, T = signal transduction mechanisms, M = cell wall/membrane, N = cell motility, W = extracellular structures, U = intracellular trafficking, secretion and vesicular transport, O = post‐translational modifications, C = energy production and conversion, G = carbohydrate transport and metabolism, E = amino acid transport and metabolism, F = nucleotide transport and metabolism, H = coenzyme transport and metabolism, I = lipid transport and metabolism, P = inorganic ion transport and metabolism, Q = secondary metabolites biosynthesis, transport and catabolism and RS = poorly characterized/uncharacterized.

Figure 13.3 Percentage distribution of predicted subcellular locations for all parent proteins in the Enterobacteriaceae database (blue) versus

E. coli

K12 MG1655 (red).

Figure 13.4 An example of a peptide biomarker, from the protein YP_017248 (boxed in red) aligned against amino acid sequences for a selection of members from the B. cereus group.

Figure 13.5 Four extracted ion chromatograms and MS/MS product ions of selected native peptides (in red) and the corresponding stable isotope‐labelled standard peptides (in black) detected in

B. anthracis

protein extract. (A) Peptides MDVDMLSNR and MDVDML*SNR, (B) peptides AIGAELDQLVK and AIGAEL*DQLVK, (C) peptides LVSIGELQPDGNR and LVSIGEL*QPDGNR, and (D) peptides SADLVQGLVDDAVEK and SADLVQGL*VDDAVEK.

Figure 13.6 Detection of peptide LVSIGELQPDGNR from 0.1 μg of protein extract from

B. anthracis

NCTC109 spiked into a background of 9.90 μg of protein extract from

B. cereus

(

B. cereus

,

B. thuringiensis

,

B. weihenstephanensis

and

B. mycoides

). (A) No selected peptide (LVSIGELQPDGNR) is detected in background of

B. cereus

; (B) Peptide LVSIGELQPDGNR is detected with a signal‐to‐noise ratio of 75; and (C) MS spectrum of peptide LVSIGELQPDGNR and its corresponding stable isotope‐labelled peptide internal standard LVSIGEL*QPDGNR; and inset (D) is the extracted ion chromatogram of the internal standard peptide.

Chapter 14

Figure 14.1 Spots picked for in‐gel digestion and identification on the strain 027 SM reference map. Surface and virulence proteins are highlighted in red. The two spots corresponding to the S‐layer protein (SlpA) are boxed (spots A1 and F6). A post‐translational cleavage event creates separate high‐molecular‐weight and low‐molecular‐weight S‐layer proteins. The series of spots corresponding to flagellin are also boxed (M2‐M7). Here, post‐translational glycosylation of flagellin leads to a series of protein spots which migrate differently.

Figure 14.2 A DIGE gel image of strain 027 SM (Cy5, red) compared to strain B‐1 (Cy3, green). The Cy2 channel (blue) is the internal standard containing a mix of protein extracts from all strains. The yellow boxes 1 and 2 highlight protein spots corresponding to surface proteins (Cwp2, SlpA) present in strain 027 SM but not strain B‐1. Yellow box 3 highlights a series of spots identified as the same protein (flagellin) but modified post‐translationally. The series of Cy3 and Cy5 labelled proteins are clearly visible as separate spots which have migrated differently within the gel, due to differences in the post‐translational modification of this protein between the different strains. Yellow box 4 highlights the SlpA high‐molecular‐weight (HMW) protein, which shows differential migration between the strains. (Chilton

et al.

, 2014.)

Figure 14.3 Proteins spots with significantly higher concentration in strain 027 SM. The proteins identified as up‐regulated in strain 027 SM by correlation analysis were matched to the ‘picking gel’ used to create the strain A reference map. The numbers indicate the rank of the protein, with protein 1 showing the greatest difference between the strains.

Figure 14.4 The

Clostrdium difficile

pathogenicity locus containing the two toxin genes (TcdB and TcdA), a σ‐factor (TcdR) thought to positively regulate transcription of the toxin genes, a possible holin (TcdE) and a potential negative transcription regulator TcdC.

Figure 14.5 Mauve analysis comparing the organization of the SlpA operon for strains 027 SM, B‐1, Tra5/5 and the 630 reference strain. Yellow bars represent areas of genetic homology between the strains, with the height of the bar representing the level of homology and ORFs being designated by bars and arrows below. The SlpA gene is shown in red in all strains, and highlighted in green in strain 630, and the gene encoding Cwp2 is shown in orange. Hypothetical or putative ORFs are shown in grey. Arrows above the genes show the direction of transcription; blue arrows indicate that the corresponding protein was detected in this analysis. CSP denotes genes encoding cell surface proteins (Slp homologues). In strain B‐1, this genetic locus shows considerable differences from the other three genomes, with an insertion of approximately 58 kb. This insertion contains 50 ORFs, the majority of being putative and uncharacterized.

Chapter 15

Figure 15.1 Peptide mass fingerprint of protein band corresponding to molecular weight of 29 kDa on SDS PAGE extracted from an ampicillin‐resistant

E. coli

.

Figure 15.2 Schematic representation of selection and fragmentation of molecular ion, selection and detection of daughter ion with the use of triple quadrupole mass spectrometer.

Figure 15.3 MRM chromatograms of mixture of ampicillin (10 mg/ml), cefotaxime (0.5 mg/ml), meropenem (0.5 mg/ml) and oxacillin (0.5 mg/ml) incubated at 37°C for 2 h (A) in phosphate buffer (B) after incubation with cell lysate of

P. aeruginosa.

The samples were diluted 1:22 times prior to injection. Chromatograms were acquired using an API3000 triple quadrupole mass spectrometer fitted with an electrospray ionization source and operated in positive ion mode.

Chapter 16

Figure 16.1 Illustration showing the distinction between microbial “identification” and “typing”, that is, what levels of resolution are needed for which levels of taxonomy. Also shown are different commonly used terms for different subgroup types.

Figure 16.2 An example of a typical spectra obtained by MALDI‐TOF‐MS analysis of

Staphylococcus aureus

. The intensities of the peaks are normalized according to the highest peak (100%), and the mass‐to‐charge ratio (m/z) is indicated at the major peaks.

Figure 16.3 A dendrogram derived from MALDI‐TOF MS spectra showing the relative similarities of

Staphylococcus

species. This demonstrates the ability to differentiate and identify

S. aureus

strains within the genus

Staphylococcus

, that is, species‐level identification, and the inability, using standard MALDI‐TOF MS protocols, to differentiate MRSA (shown by *) from MSSA strains (shown by **).

Figure 16.4 A dendrogram derived from MALDI‐TOF MS spectra of the relative similarities of

S. aureus

strains. The subtypes identified by pulsed‐field gel eletrophoresis (PFGE) are discriminated in the separate subclusters, indicating a high correlation between the different approaches. This shows that MALDI‐TOF MS protocols have potential for strain‐level differentiation in

S. aureus

.

Figure 16.5 Top‐down and bottom‐up approaches employed for proteotyping. In top‐down approaches (left part of the figure), the intact proteins are separated by liquid chromatography without prior digestion. Note the highly charged states of the intact proteins after ionization. Here, the generated tandem MS spectra reflect the mass differences of intact peptides. In bottom‐up approaches (right part of the figure), the proteins are first digested into peptides (for example, using a digestive enzyme, such as trypsin), whereby the peptides subsequently are separated by LC and ionized prior to mass analysis. Here, the tandem MS spectra reflect the mass differences of amino acids.

Figure 16.6 Workflow of proteotyping. The

Helicobacter pylori

strain ATCC 26695 was analyzed with the bottom‐up proteotyping approach. Step 1 shows sample preparation protocols involving cell fractionation protocols, or methods for keeping cells intact are employed. Step 2 is the digestion of the sample proteins into peptides. Step 3 represents the analysis of the generated peptides, using LC‐MS/MS. An example of nine peptides coming from the protein thioredoxin reductase of

H. pylori

is shown. In Step 4, the peptides are aligned against available genomes in databases, here exemplified by two sequences for the thioredoxin reductase, one from the strain

H. pylori

ATCC 26695 and one from the strain

H. pylori

J99. The green‐shaded areas represent the identified peptides. The yellow‐marked amino acids show where the amino acid sequences differ between the two strains. The red arrows show where the identified peptides contain these amino acid differences, thus enabling identification of the correct strain, as these are unique for the particular analyzed strain,

H. pylori

ATCC 26695.

Figure 16.7 A general overview of the proteotyping workflow. The process starts in the top‐left corner with sample preparation and finishes with the proteotyping results in the lower‐right corner. Each box represents a specific experimental or analytical stage of the workflow. The three cylinders correspond to reference databases that are essential for the analysis, including a comprehensive protein database, a set of reference genomes and their association with the taxonomic tree. Four fundamental bioinformatics stages are highlighted by the dashed rectangle in the centre of the graph. The final proteotyping result is an estimate of the taxonomic composition of the sample.

Figure 16.8 The lowest common ancestor (LCA) algorithm. The figure shows three examples (A, B, C) of peptides with matches to reference genomes at different levels in the taxonomic tree. (A) A single peptide has matched nodes 1 and 2. These nodes have a lowest common ancestor on the genus level, which means that this peptide is discriminative on the genus level, but not on any lower level. (B) A single peptide has matched nodes 1 and 2, leading to a lowest common ancestor above the genus level. Such a peptide contains very little useful information for proteotyping on species and subspecies levels. (C) Three different peptide have matched nodes 1, 2 and 3. There is no lowest common ancestor; each peptide is discriminative on its own at the level it matched (species, subspecies and genus, respectively).

Figure 16.9 The number of discriminative peptides decreases rapidly at lower taxonomic levels. This figure shows the number of discriminative fragments and the family, genus and species levels for the bacteria,

E. coli, S. aureus

and

S. pneumoniae

. The drop in discriminative fragments for each taxonomic level differs between species and is dependent on the comprehensiveness of the reference databases and the similarity of the proteome of the closest relative. The numbers were calculated on the basis of an implementation of the workflow using X!! Tandem (Bjornson

et al.

, 2008) together with a reference databases composed of NCBI GenBank non‐redundant proteins (Benson

et al.

, 1999) and the Human Microbiome Project (HMP) reference proteomes (Turnbaugh

et al.

, 2007). Alignment of peptides were done using BLAT (Kent, 2002) against a manually curated version of NCBI RefSeq (Pruitt, Tatusova, and Maglott, 2007) and HMP reference genomes (Turnbaugh

et al.

, 2007).

Figure 16.10 Numbers of strain‐specific peptides from different mixtures of two

H. pylori

strains (J99 and ATCC 26695). The x‐axis shows the how the strains were mixed, ranking from 9:1 to 1:9. Proteotyping analysis revealed that <100 peptides were discriminative, which corresponds to ~10% of the total identified peptides

Figure 16.11 Proteotyping protocol implemented directly on nasopharyngeal samples. (A) A clinical sample identified as positive for

Moraxella catarrhalis

, by culture and MALDI‐TOF‐MS identification analysis. The proteotyping workflow found 34 unique peptides for

M. catarrhalis

and showed that this was the dominant species in the sample. (B) Sample identified as positive for

Staphylococcus aureus

by culture and MALDI‐TOF‐MS identification analysis. The proteotyping workflow found 26 unique peptides for

S. aureus

as the dominant species in the sample.

Figure 16.12 Methodologies for phenotypic and genotypic characterizations. The taxonomic resolution for the different methodologies are indicated by a solid black line; the dotted lines indicate applications to limited extents. Some methodologies, such as LC‐MS/MS proteotyping, are applicable for resolving microorganisms over the entire taxonomic range indicated, that is, from the family to strain levels.

Chapter 17

Figure 17.1 Cystic fibrosis lung infections caused by different species of pathogens. Whereas

S. aureus

is predominant in the early stages of infections,

P. aeruginosa

is the most dominant pathogen in late stages of infections. Figure adapted from the 2009 Patient Registry Report, issued by the Cystic Fibrosis Foundation, with permission [23].

Figure 17.2 Sampling points of isolates and life expectancy of patients.

P. aeruginosa

isolates were isolated from 11 CF patients over 35 years. Each isolate is labelled by different symbols, whereas patient life expectancy is shown with grey bars.

Figure 17.3 Hierarchical clustering of the RNA transcripts using Euclidean distances. The red block represents an increase in signal, whereas the green block shows a reduction in signal. Figure adapted from Yang, L.

et al.

, 2011. Bacterial adaptation during chronic infection revealed by independent component analysis of transcriptomic data.

BMC Microbiology

11

, 184 (2011)10.1186/1471‐2180‐11‐184) [165].

Figure 17.4 Iron acquisition systems employed by

P. aeruginosa

. Siderophore‐based systems including

pvd

[220,221](A) and

pch

[222](B) are both involved in Fe

3+

uptake;

phz

(C) is involved in Fe

2+

uptake [223]; heme‐acquisition‐based systems including

phu

(D) and

has

(E), are involved in uptake of hemeproteins.

Chapter 18

Figure 18.1 Bottom‐up versus top‐down proteomics.

Figure 18.2 Targeted versus discovery mode in top‐down proteomics.

Figure 18.3 (A)

Neisseria meningitidis

colonies on an epithelial cell, (B)

N. meningitidis

bacterium and its type IV pili, (C) modeling of a type IV pilus, (D) PilE protein: the major component of type IV pili.*

Figure 18.4 High‐resolution FT‐ICR mass profile of PilE alongside a model showing all PTMs (orange) on the protein structure. Once all PTMs are taken into account, the experimental spectrum (black) correlates very well with a theoretically generated isotope pattern (red).

18

Figure 18.5 Comparison of the total ion chromatograms obtained for the LC‐MS analysis of intact proteins extracted from

S. sonnei

and

E. coli

lysates.

Figure 18.6 Sequence coverage (

c/z

ions) obtained for the 50S L7/L12 ribosomal protein of mass 12,198.5 Da (using ETD fragmentation).

Chapter 19

Figure 19.1 Q Exactive instrument layout. The instrument incorporates an S‐lens, a selective quadrupole mass filter, a HCD collision cell interfaced to the C‐trap, and a high‐resolution Orbitrap mass analyzer. The figure is reprinted with permission from Thermo Fisher Scientific.

Figure 19.2 SDS‐PAGE analysis of biotin‐enriched surface proteins from fecal bacterial populations. (A) Comparison of biotin‐enriched surface proteins and proteins in a whole‐cell lysate of bacterial cells separated from a fecal sample by differential centrifugation. (B) Comparison of biotin‐enriched surface proteins of bacteria from fecal samples collected from three healthy individuals (1–3). From donor 1, a sample (1*) was also collected one month after the first sample. For each sample, one example of three technical replicates treated separately is shown. Molecular weights (kDa) of standard proteins are indicated.

Figure 19.3 Origin of the proteins detected by the cell surface protein profiling method in a fecal sample. Ratios of proteins from bacteria, eukaryotes, viruses, and archaea detected after (A) in‐gel trypsin digestion or (B) in‐solution trypsin digestion are shown.

Figure 19.4 Principal component analysis of surface protein profiles of fecal bacteria. The bacteria were isolated from three or more replicate samples collected from three different individuals (1, black circles; 2, gray circles; 3, black squares). Biotin‐enriched proteins were analyzed by RP‐LC‐MSMS as three technical replicates.

Chapter 20

Figure 20.1

Proteogenomics map.

The previously annotated genes are indicated with arrows, whereas the peptides established by tandem mass spectrometry are indicated with vertical bars. These peptides are mapped onto the nucleotide sequences. The different outcomes in terms of re‐annotation of the genome are indicated.

Figure 20.2

Re‐annotation of the translation initiation codon for the

dnaA

gene in

Deinococcus deserti

genome.

The MS/MS spectrum corresponding to the SQEIWADVLGYVR peptide labeled with the TMPP reagent has been recorded with an LTQ‐Orbitrap XL instrument (Thermo Scientific). This spectrum enables this peptide to be established as the N‐terminus of the matured protein. It can be explained by a translation start at the ATC initiation codon highlighted in bold red in the genome sequence. The corresponding polypeptide sequence and the previous annotations are shown in red and black, respectively, for three different

Deinococcus

genomes.

Figure 20.3

Proteogenomics strategy using a draft genome sequence applied to the

Tistlia consotensis

bacterium.

Tistlia consotensis

cells were grown in three different media differing in sodium chloride concentration. For each condition, three biological replicates were carried out. Cells were collected by centrifugation and subjected to whole‐cell shotgun proteomics. To interpret the tandem mass spectrometry results, the genome was sequenced, resulting in a draft genome comprising 2377 contigs. These were translated in the six possible ways to establish a comprehensive database of 52,246 open reading frames (ORFs). The interpretation resulted in the identification of 4686 unique peptides and the abundance comparison of 872 proteins.

Chapter 21-1

Figure 21A.1 Example of effect of downsampling on part of a spectral profile. The original profile with all sampled points is seen in (A). After downsampling, fewer points remain (B) while maintaining the shape of the peak.

Figure 21A.2 Demonstration of too aggressive downsampling setting. The original profiles in (A) contains three peaks. After downsampling, the profile in (B) only contains one peak.

Figure 21A.3 The result of different baseline subtraction algorithms: (A) the binned baseline, (B) the monotone minimum, (C) the moving bar, and (D) the rolling disk. The top green curve is the raw profile, and the bottom red curve is after background subtraction.

Figure 21A.4 Result of the different curve smoothing algorithms: (A) Savitsky–Golay, (B) Gaussian, (C) moving average, and (D) Kaiser. All algorithms were applied to the profile after background subtraction with a moving disk algorithm. The top green profile is the raw profile, and the bottom blue is the smoothed profile.

Figure 21A.5 The results of two different preprocessing workflows: (A) a moving average followed by peak detection with local maxima and (B) a Kaiser window smoothing followed by a CWT peak analysis. The top green profile is the original profile, and the bottom orange is the smoothed profile. Small circles on the bottom curve indicate the presence of a peak.

Figure 21A.6 Average spectral profile (red) derived from the member profiles (gray). Peaks are only retained if they are detected in the majority of members. Adjacent peaks are marked with different colored lines.

Figure 21A.7 Examples of profiles yielding (A) high and (B) very low pairwise similarity values using the Pearson correlation coefficient while containing the same peaks.

Figure 21A.8 Examples of profiles which are affected by a different choice of intensity threshold for peak detection when using a binary coefficient and which also score differently with binary versus curve‐based coefficients.

Figure 21A.9 Illustration of the importance of the position tolerance settings when using a binary similarity measure. The middle peaks will be considered as matching only when the position tolerance will be substantially increased.

Figure 21A.10 An illustration of the differences of the different cluster algorithms: on top, the original similarity matrix; (A) the averaged UPGMA reduction; (B) the single linkage approach, and (C) the complete linkage method. All methods yield a slightly different merged similarity matrix. The cells on which the merged matrix is based are marked with the same color in the original and merged matrix for entry C.

Figure 21A.11 Dual two‐dimensional plot of a PCA analysis. Left: a plot of the samples, and right: a plot of the corresponding peak classes (see text for further explanation). The color of the samples is based on the genus.

Figure 21A.12 (A) Details of a peak matching analysis with (B) the resulting peak matching table.

Figure 21A.13 An example of a three‐dimensional PCA plot. Samples are colored as in Figure 12.