Use of MALDI-TOF Mass Spectrometry and a Custom Database to Characterize Bacteria Indigenous to a Unique Cave Environment (Kartchner Caverns, AZ, USA)

Podsumowanie

This work details procedures for rapid identification of bacteria using MALDI-TOF MS. The identification procedures include spectrum acquisition, database construction, and follow up analyses. Two identification methods, similarity coefficient-based and biomarker-based methods, are presented.

Streszczenie

MALDI-TOF mass spectrometry has been shown to be a rapid and reliable tool for identification of bacteria at the genus and species, and in some cases, strain levels. Commercially available and open source software tools have been developed to facilitate identification; however, no universal/standardized data analysis pipeline has been described in the literature. Here, we provide a comprehensive and detailed demonstration of bacterial identification procedures using a MALDI-TOF mass spectrometer. Mass spectra were collected from 15 diverse bacteria isolated from Kartchner Caverns, AZ, USA, and identified by 16S rDNA sequencing. Databases were constructed in BioNumerics 7.1. Follow-up analyses of mass spectra were performed, including cluster analyses, peak matching, and statistical analyses. Identification was performed using blind-coded samples randomly selected from these 15 bacteria. Two identification methods are presented: similarity coefficient-based and biomarker-based methods. Results show that both identification methods can identify the bacteria to the species level.

Wprowadzenie

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) has been shown to be a rapid and reliable tool for identification of bacteria at the genus, species, and in some cases, strain levels^1-4. MALDI-TOF MS ionizes biological molecules (typically proteins) that originate from cell surfaces, intracellular membranes, and ribosomes from bacterial whole cells or protein extracts^1,5. The resulting peaks form characteristic patterns or “fingerprints” of the bacteria analyzed¹. Identification of bacteria is based on these mass-to-charge “fingerprints”.

Two of the most commonly used identification strategies are library-based and bioinformatics-based strategies¹. Library-based approaches involve comparing the mass spectra of unknowns to previously collected mass spectra of known bacteria in databases/libraries for identification. Commercially available software, such as BioNumerics, Biotyper, and SARAMIS software packages, as well as open source software tools, such as SpectraBank⁶, are available to facilitate the comparison and quantification of similarity between mass spectra of unknowns and reference bacteria. Bioinformatics-based approaches usually rely on fully sequenced genomes of bacteria for identification. In contrast to library-based approaches which do not involve identification of the biological nature of particular peaks, bioinformatics-based approaches involve protein identification¹.

The majority of recent MALDI fingerprint-based studies have used library-based approaches to identify bacteria¹. Library-based approaches require construction of databases and comparison of the similarity between mass spectra. Studies show that many experimental procedures, such as medium^3,7, cultivation time⁸, sample preparation method³, and matrix used⁹, affect the mass spectra obtained. Furthermore, some closely-related species and strains generate spectra with only subtle differences. Thus, library-based approaches require rigorously standardized procedures to generate highly reproducible mass spectra between replicates. Minor variations in protocols may compromise the efficacy of identification, especially at the subspecies and strain levels^1,3,10. However, neither manufacturer-provided reference databases nor reported custom databases include visually documented procedures for database construction and/or application of a data analysis pipeline. For this reason, the objective of this work was to develop, apply, and demonstrate a comprehensive and detailed procedure for library-based bacterial identification using MALDI-TOF MS.

In this demonstration, mass spectra of 15 bacteria isolated from a karstic environment (Kartchner Cavern, AZ, USA) were collected and imported into software to construct a model database. Data processing and the analysis pipeline were detailed using the model database. Finally, mass spectra of blind-coded bacteria which were randomly selected from these 15 bacteria were collected again and compared to the reference spectra in the model database for identification. Results show that bacteria can be correctly identified either based on similarity coefficients or potential biomarkers/peak classes.

Protokół

Caution: Unidentified bacteria from any environment may be pathogenic and must be handled with caution using appropriate biosafety protocols. Work with live cultures must be performed in a Class II biosafety cabinet using Biological Safety Level 2 (BSL-2) procedures. More information about BSL-2 procedures is available in the CDC/NIH manual titled, "Biosafety in Microbiological and Biomedical Laboratories," pages 33-38. The document is available online at http://www.cdc.gov/biosafety/publications/bmbl5/BMBL.pdf. Appropriate personal protective equipment (PPE), including lab coats/gowns, safety glasses, and nitrile or latex gloves, must be worn. Standard microbiological practices and precautions must be followed, and biohazardous waste must be discarded appropriately.

Bacteria used in this demonstration were isolated from Kartchner Caverns, AZ, USA, from four environments, including dry speleothem, flow stone, moist speleothem and stalactite drip (Table 1). All isolates were identified by 16S rDNA sequencing and kept at -80 °C in 25% glycerol-R₂B medium. All experiments were completed at RT.

Note: We recommend using the same sample preparation method to acquire mass spectra for database construction and mass spectra of unknowns. Sample preparation method has been shown previously to affect spectrum quality and reproducibility³. Using a different sample preparation method may cause incorrect identification of unknowns, especially when higher taxonomic resolution (e.g., at the strain level) is desired.

1. Deposition on the MALDI Target

Caution: Several protocols to obtain protein extracts require use of acids and organic solvents that must be utilized in accordance with guidelines and information contained in their respective Materials Safety Data Sheets (MSDS). Appropriate PPE must be worn and will vary based upon type and volume of chemicals used (e.g., lab coats/gowns, gloves, safety glasses, and respiratory protection must be used when working with significant quantities of toxic, flammable solvents, such as acetonitrile, and corrosive acids, such as formic and trifluoroacetic acids).

1. Deposit 1 µl protein extract containing no viable cells (obtained using appropriate, previously described protocols^11-13) onto a stainless steel MALDI target plate and allow it to dry. Overlay the dried protein extract with 1 µl matrix solution (α-cyano-4-hydroxy-cinnamic acid solution), and allow it to dry.
2. For each biological replicate, spot an appropriate number of technical replicates (5 to 20 technical replicates). Here, spot 10 technical replicates for each biological replicate and 3 biological replicates were prepared.
   Note: We recommend using a polished MALDI steel target plate when using the protein extraction sample preparation method. Using ground steel target plates may cause spreading and unintentional mixing of the different samples outside of individual sample wells.
3. Deposit 1 µl calibrant standard onto the target plate and allow it to dry. Overlay with 1 µl matrix solution and allow it to dry.
4. Deposit 2 µl matrix solution onto the target plate as a negative control.

2. Mass Spectra Acquisition

1. Use a MALDI-TOF mass spectrometer equipped with a nitrogen laser (λ = 337 nm) and operated using Bruker FlexControl software.
2. Collect each mass spectrum in positive linear mode by accumulation of 500 laser shots in 100 shot increments. Set the ion source 1 voltage to 20 kV; ion source 2 voltage to 18.15 kV; and the lens voltage to 9.05 kV. Note that these parameters are instrument-specific and might require adjustment on other instruments to obtain optimal results.
3. Set the mass-to-charge range for automated spectrum evaluation from 2 to 20 kDa per charge. Use the centroid peak detection algorithm. Set the minimum resolution threshold at 100 Da. Set the signal to noise ratio (S:N) threshold at 2. Set the minimum intensity threshold at 100.

3. Database Construction

1. Database design
   1. Create a new database in BioNumerics 7.1 using the "New database wizard".
   2. Create a spectrum experiment type, e.g., Maldi, using commands in the "Experiment Types" panel.
   3. Create the levels using the "Database design panel". Add new levels using the "Level > Add new level…" command in the "Database" menu. Here, create “Species” level, “Biological replicate” level” and “Technical replicate” level, respectively.
2. Importing and preprocessing raw mass spectra
   1. Export the raw mass spectra as .txt files using FlexAnalysis by clicking the "Export > Mass spectrum" command in the "File" menu.
   2. Import the raw mass spectra (.txt files) into the database in the level of the technical replicates.
   3. Preprocess the raw mass spectra.
      1. Import and resample (using a quadratic fitting algorithm).
      2. Perform a baseline subtraction (with a rolling disc with a size of 50 points).
      3. Compute noise [Continuous Wavelet Transformation (CWT)], smooth (Kaiser Window with a window size of 20 points and beta of 10 points), and perform a second baseline subtraction (rolling disc with size of 200 points).
      4. Detect peaks [CWT with a minimum signal to noise ratio (S:N) of 10].
   4. After preprocessing, save characteristic patterns of each mass spectrum, such as peak lists containing peak sizes, peak intensities, S:N, etc., in the database.
3. Creating composite mass spectra
   1. Create composite spectra from preprocessed spectra using the "Summarize…"command in the "Analysis" menu. Choose the "Biological replicate" as target level.
   2. Here, combine mass spectra of 10 technical replicates of the same colony to yield a composite mass spectrum for that colony, resulting in three composite mass spectra for that isolate at the "Biological replicate" level.
   3. Here, summarize the three composite spectra to create one composite spectrum for that isolate at the "Species" level.
      Note: The composite spectrum is the point-by-point average of the technical replicates. Replicates with a similarity (Pearson correlation) to the average of lower than 95% (default setting) are excluded from the composite. Peaks on the composite spectra are only called if they are present in 75% (default setting) of the included replicates. For the biological replicates, these settings were 90 and 60%, respectively.

4. Mass Spectrum Data Analysis

1. Select the entries in the database and create a comparison by clicking the "Create new comparison" command in the "Comparisons" panel.
2. Here, use the mass spectra at the "Technical replicate" and/or "Biological replicate" levels to show comparisons and analyses.
3. Similarity-based cluster analysis and multi-dimensional scaling (MDS)
   1. Create groups with colors. Select the three biological composite mass spectra and click the "Create new group from selection" command in the "Groups" menu to create a group for the corresponding isolate. Designate a color automatically used for these three mass spectra.
   2. Alternatively, define field states with corresponding colors using the commands in the "Database entries" panel such that any grouping based on this defined field uses the same color defined for this group.
   3. Perform cluster analysis. Click the "Calculate cluster analysis" command in the "Clustering" menu. On the Comparison settings Page 1, select the "Pearson correlation" and leave other parameters as default. On page 2, select "UPGMA". Then click "Finish".
   4. Obtain an MDS plot using the "Multi-dimensional scaling…" command in the "Statistics" menu.
4. Peak matching
   1. Click on the spectrum type "Maldi" in the "Experiments" panel. Then select "Layout > Show image". Spectra are shown as gel bands.
   2. Perform peak matching using the "Do peak matching" command in the "Spectra" menu.
5. Identification of peak classes
   1. Perform principal component analysis (PCA). Highlight the "Maldi" experiment type in the "Experimental" panel and use the "Principal Components Analysis…" command in the "Statistical" menu to perform PCA.
   2. Perform two-way clustering. Click the "Statistics > Matrix mining"… in the "Comparison" window. The intensity of the peaks matched to the peak classes is represented using different colors (heat map).

5. Bacteria Identification with a Custom Database

1. Similarity coefficient-based method
   1. Create a comparison and generate a dendrogram based on mass spectra at the "Technical replicate" level as described in step 4.3.3. Save the dendrogram for comparison of similarity.
   2. Select an unknown mass spectrum, and click the "Analysis > Identify selected entries". The identification dialog box appears.
   3. Select the "Comparison based" classifier type (or a stored classifier) and click "next". On the next page, choose the saved dendrogram as a reference comparison and then click "next".
   4. Choose the "Basic similarity" as an identification method and then click "next".
   5. Choose the "Maximum similarity" as scoring method. Type in appropriate threshold values and minimum difference values for each parameter and then click "next".
   6. Once the calculations are completed, the identification window appears. In the "Results" panel, the members of the database that best match the unknown are listed.
   7. Save the identification project and validate the identification using the "cross-validation analysis" command in the "Identification project" panel.
2. Potential biomarker-based method
   1. Define peak classes. In the "Matrix Mining" window, select sets of peaks sharing common characteristics and define these peaks as specific peak classes (potential biomarkers) using "Spectra > Manage peak class types…" in the "Comparison window".
   2. Here, define peak classes specific for each isolate for all the 15 isolates.
   3. Select the mass spectra of unknowns and match the peaks of these spectra to the defined peak classes as previously described.

Wyniki

The databases constructed in this demonstration had four levels, from highest to lowest level, including “All levels”, “Species”, “Biological replicate” and “Technical replicate”, respectively (Figure 1A). The “Technical replicate” level contained all the preprocessed spectra of technical replicates. The “Biological replicate” and “Species” levels contained the composite (summary) spectra. “All levels” contain...

Dyskusje

This demonstration showed detailed procedures of characterization and identification of bacteria using MALDI-TOF MS and a custom database. In comparison to traditional molecular methods, for example, 16S rDNA sequencing, MALDI-TOF MS-based fingerprint methods facilitate more rapid identification of diverse bacteria. Because of its robustness, this technique is widely used to characterize bacteria, viruses, fungi and yeast from the environment and in clinical settings^1,14-16. Moreover, MALDI-TOF MS has been rep...

Materiały

Name	Company	Catalog Number	Comments
α-cyano-4-hydroxy-cinnamic acid	ACROS Organics	163440050	≥ 97%, CAS 28168-41-8
Bruker FlexControl software	Bruker Daltonics		version 3.0
Bruker FlexAnalysis software	Bruker Daltonics		version 3.0
Bionumerics software	Applied Maths		version 7.1