1. Finding a Needle in a Haystack: Using High Resolution Mass Spectrometry in Targeted and Non Targeted Searching for Food Contaminants
Erik Verschuuren
Strategic Account Manager, Food Safety and Environmental
Thermo Fisher Scientific

2. Where Target Analysis has Failed?
Melamine in milk powder
Illegal dyes in chilli powder
Acrylamide in crisp bread & French fries
BADGE in canned fish
2-Isopropylthioxanthone (ITX) in packaged foods
Illegal addition of proteins to milk
Industrial chemicals & pharmaceuticals in environment
So where the targeted analysis failed: few examples, the use of illegal proteins (kolagene from leather waste) being the recent one.
The reason: cost cutting in most cases (melamine, dyes, protein…)
The others are unexpected contamination issues like the residues of inks and various modifiers in the cardboards, acrylamide in crisps and many others…

3. Different Modes of Analysis
Pure Qualitative
Pure Quantitative
Detect & Quantify
Determine structure
Structural ID
Compound Confirmation
Reaction Monitoring
Process Monitoring
Metabolism
Proteomics
Metabolomics
Targeted Analysis
Non-Target Analysis
Screening
Quantification
Confirmation
Profiling
Fingerprinting, Statistical
Database searching
Identification of unknown
Find your needles

4. Triple Quadrupoles Used for Targeted Analysis
TSQ Quantum Access Max
TSQ Quantum Access
TSQ Quantum Ultra
TSQ Vantage
The following is what Thermo Scientific has to offer in the TSQ line. For the ultimate value we have the TSQ Quantum Access Max. For the utmost detection needed TSQ Vantage which was introduce in 2008.
Ultimate Value
Ultimate Performance

5. However…. Technology is evolving… ….direction glaring in the mist

6. Shift Towards HRAM : Crossover Point
Official Food Control Authority of the Canton of Zurich

7. Shift Towards HRAM : Crossover Point - cont

8. High Resolution Analyzers Are Claiming Their Share
Exactive
QExactive
LTQ Orbitrap
Bench-top HRMS
Ultimate Performance

9. How Much Mass Accuracy is Needed?
340
350
360
370
380
390
400
410
m/z 10 20 30 40 50 60 70 80 90
100
Relative Abundance C 27 H 11 O 2 N
[M+H]
Mass accuracy
Number of hits
100 ppm 34
50 ppm 17
30 ppm 11
10 ppm 4
5 ppm 2
3 ppm
Compounds containing
C[5-15]N[0-10]O[0-15]H[0-40]

10. Selectivity Increases With Higher Mass Accuracy
± 0.1µ
± 0.05µ
± µ

11. Detection of Isobaric Compounds in Mixtures
Element Exact Mass H C N O
Is a simultaneous measurement possible?
289.00
289.05
289.10
289.15
289.20
m/z 10 20 30 40 50 60 70 80 90
100
Relative Abundance
R = 10,000
289.00
289.05
289.10
289.15
289.20
m/z 10 20 30 40 50 60 70 80 90
100
Relative Abundance
R = 20,000
289.00
289.05
289.10
289.15
289.20
m/z 10 20 30 40 50 60 70 80 90
100
Relative Abundance
R = 100,000
C9H21O2P1S3
Terbufos
C13H21O3P1S1
Iprobenfos
C15H17N4Cl1
Myclobutanil
C11H20N4O3S1
Epronaz
C11H21N4O3P1
Pirimethaphos
C16H20N2O3
Imazamethabenz
How would this be used for residue analysis in food? These six pesticides have the same nominal mass, but different exact mass. The six compounds are isobaric compounds. Is a simultaneous measurement possible for the six compounds? At 20,000 resolution, which TOF instrument can reach, impossible. But at 100,000 resolution, total separation, which TOF unable to reach.
0.1033µ 11

12. Detection of Analytes in Heavy Matrix
Pesticide Mixture in Horse Feed Matrix
239.00
239.05
239.10
239.15
239.20
239.25
m/z 20 40 60 80
100
Relative Abundance C 11 H 19 O 2 N 4
6.50 ppm
R = 15,000
Error = 6.50 ppm
R = 80,000
Error = 0.32 ppm
239.00
239.05
239.10
239.15
239.20
239.25
m/z 20 40 60 80
100
Relative Abundance C 11 H 19 O 2 N 4
0.32 ppm

13. Targeted Screen Report
Target Screen Approach
Raw Data
Targeted Screen Report
This slide shows the main targeted screen approach:
GC or LC HRMS data acquisition (non targeted)
Analysis of raw data by searching against database of known compounds (ExactFinder)
Library of “Knowns”

14. Non Targeted Analysis – What’s the challenge?
High resolution MS data acquisition - Unknown is unknown… - Matrix presence - Variability of samples - Small concentrations of potential threats - Extraction procedures: polarity, pH
Data processing?

15. MS Detectors for Unknowns
Ion Trap MS
High Resolution
Orbitrap MS
For non-target screening, ion trap MS can be applied.
Ion Trap MS ID peaks of unknowns by full scan mode. Confirmation by MSn in a second run.
More often, HRMS are used because of the accurate mass and high resolution offered by these instruments.
The same, when use HRMS for non-target screening, analytes defined after data acquisition
↑Selectivity by exact mass
Confirmation by accurate mass & ion ratio deviation
Quantitation is available for the instrument with good dynamic range, such as Orbitrap Exactive.
Click
Full scan mode acquisition
Confirmation by MSn
Exact mass and high resolution data
Confirmation by isotopic pattern and MS/MS
Confirmation and Quantitation in 1 run

16. Key Requirements: Mass Accuracy + Resolution
m/z
C, N, O, H
ppm error
Number of hits
150 11 50 4 10 1 5 2
m/z
C, N, O, H
ppm error
Number of hits
150 38 50 12 10 3 5 2
Mass accuracy and resolution are important factors but strongly depend on the mass... See example of two ions, and number of potential hits based on various mass errors: 10 ppm with mass 127 means 1 hit, while it means 3 hits with mass 700. And gets worse when other elements are considered.
Conclusion: the mass accuracy and resolution is important but needs additional information.

17. Example: Analysis of Milk Powder
Milk sample 2g
Add 20 ml Acetonitrile/Water
70:30 v/v
Extraction 20 min
Centrifugation rpm
10 minutes
Filtration through
0.2 µm PTFE into LC vial
UHPLC/Exactive MS analysis
SIEVE data processing
This is an example of the attempt to use a simplistic sample preparation approach in conjunction with HRMS to search for “unknowns” in milk powder. Various contaminants (pesticide, dyes, adulterants) were spiked into milk and processed as shown above. The results obtained are described below.
Poster: NEW STRATEGIES FOR FOOD PROFILING AND MONITORING USING HIGH RESOLUTION MASS SPECTROMETRY APPROACH, M. GODULA1, A. CHARLTON2, D. ROBERTS2 AND K. MITTENDORF1

18. TIC and BPC of Milk Extract - No Difference…
TIC (Total Ion Current)
BPC (Base Peak Chromatogram)
The most critical issue in unknown analysis is to find a presence of a “foreign” peak. However by looking on the TIC or base peak chromatogram (BPC) we can see some peaks but certainly not to say if they are natural of contaminants.
The key question is: where is the information?

19. Automated Component Detection - ExactFinder
Background subtracted file
Components extracted
Some of the new data processing algorithms like to automated component detection in ExactFinder can reduce the data complexity, remove background and show the presence of a chromatographic peak: picture below shows clearer chromatogram when data reduction is performed but info is still limited - too many unknowns. Now it is important to decide what is the peak we look for.

20. Comparing Contaminated Sample (10ppm) With Blank
The only visually identified difference
Fenthion
Sudan dyes
Melamine Cyanuric acid
Urea
We can compare the contaminated (or real) sample with a typical blank sample. When looking on the chromatograms shown above, we can find only one visible difference while more compounds are spiked into sample at reasonably high concentration of 10 ppm.
By only manual data processing it is impossible to reveal any difference between samples. We have identified only one visible difference – blue rectangle.
Let’s look at this one in more detail:
click

21. Blank Milk vs. Spiked Milk – Visual Comparison Only
20 hits < 10 ppm
CHNOSP
We can look at the peak, subtract backgroung, take spectra and what we see is a mix of peaks. Let’s assume the highest one is the pseudomolecular ion.
Large list of positive hits is generated – no confidence in the results unless more parameters are involved. We can now apply spectral searching, isotopic pattern fitting an so on and hope we have the right one!
But what about the others? We cannot even see them on the chromatogram.
click

22. SieveTM –Data Comparison Tool
Blank sample
Alignment
Framing
ChemSpider Search
Real sample
Compensation of RT changes
Table of identified frames
Table of positive hits
Sieve is the software tool originally developed for identification of metabolites in metabolomic studies. It is designed to make alignment of retention times of blank (reference) sample and real (contaminated) sample. After alignment the Sieve identifies frames within the 3D space of retention time, m/z and abundance. Each frame is unique and defined by retention time, mass and abundance. After generating list of frames, those are compared between 2 sets of data and differences are found. Those are then subjected to search of accurate masses against the web based databases. Click

23. Sieve™ – Framing Process
Explain the Sieve framing process (frame extraction) – describes how Sieve creates frames in the 3D space of the chromatographic data and classifies them according to abundance.

24. Sieve™ – Database Search Option
ChemSpider search
Web based
Number of databases available
These are new features built in the Sieve: possibility to search ChemSpider in multiple databases and for multiple molecular modifications (adducts) or use Targete MZ Lookup feature where the search through custom built database makes the search more targeted and faster: but library is necessary.
Targeted search MZ Lookup
Custom build library
Adducts searched

25. Advanced Algorithms – Elemental Composition Calculation
ExactFinder considers isotopic pattern as a confirmation criteria
MH+
Empowering confidence by elemental composition confirmation is essential. By applying isotopic pattern recognition algorithm available in ExactFinder the predicted elemental composition can be confirmed.
ASMS 2010 Poster-T501: High-Resolution LC-MS Analysis of Compounds for Early Drug Discovery: Quantitative and Qualitative Screening in One Pass, Caroline Ding1, Jim Shofstahl1, Karen Salomon1, Hans Pfaff2, and Thomas McClure1

26. Isotopic Pattern Fitting is Influenced by Resolution
The picture shown here documents one more important factor. If we want to fit the elemental composition to the isotopic pattern, resolution of the acquired data is essential. On the picture shown here the pattern looks OK but if we look at the least abundant isotope ( ) we can see by zooming in
Click
That the correct isotope is much less abundant then adjacent matrix peak. Without ultra high resolution applied the isotopic pattern would NOT fit.:
Argument for Exactive or Orbitrap.
Table of candidates
Isotopic patterns compared
Resolution is needed to separate overlapping peaks

27. Conclusions – Future View
Minimize Sample preparation
Simplified sample preparation strategies
Direct analysis where possible
Advanced data processing algorithms must be used
Data extraction and background subtraction
Comparative approaches
Libraries of accurate masses and fragments required
Isotopic pattern fitting
Validation
Validation for diverse matrices required
Self explanatory.