Forensic Applications of Gas Chromatography/Mass Spectrometry, High
Performance Liquid Chromatography—Mass Spectrometry and Desorption
Electrospray Ionization Mass Spectrometry with Chemometric Analysis
A dissertation presented to
the faculty of
the College of Arts and Sciences of Ohio University
In partial fulfillment
of the requirements for the degree
Doctor of Philosophy
Xiaobo Sun
March 2012
© 2012 Xiaobo Sun. All Rights Reserved. 2
This dissertation titled
Forensic Applications of Gas Chromatography/Mass Spectrometry, High
Performace Liquid Chromatography—Mass Spectrometry and Desorption
Electrospray Ionization Mass Spectrometry with Chemometric Analysis
by
XIAOBO SUN
has been approved for
the Department of Chemistry and Biochemistry
and the College of Arts and Sciences by
Peter B. Harrington
Professor of Chemistry and Biochemistry
Howard D. Dewald
Dean, College of Arts and Sciences 3
Abstract
SUN, XIAOBO, Ph.D., March 2012, Chemistry
Forensic Applications of Gas Chromatography/Mass Spectrometry, High
Performace Liquid Chromatography—Mass Spectrometry and Desorption
Electrospray Ionization Mass Spectrometry with Chemometric Analysis
Director of Dissertation: Peter B. Harrington
This dissertation includes three forensic applications of instrumental analysis techniques: gas chromatography/mass spectrometry (GC/MS), high performance liquid chromatography/electrospray ionization-mass spectrometry (HPLC/ESI-MS) and desorption electrospray mass spectrometry
(DESI-MS). By using these instrumental analysis techniques and chemometric analysis, jet fuel samples, illicit drugs and Panax quinquefolius
L (American ginseng) samples were analyzed rapidly and conveniently.
Fast GC was combined with fast scanning quadrupole ion trap (QIT)
MS and used to classify jet fuels for the first time by using a fuzzy rule- building expert system (FuRES) classifier. These data were pretreated with and without wavelet transformation (WT) and evaluated with respect to classification rates and 99.8±0.5% classification accuracy was obtained in both cases. Optimized-partial least squares discriminant analysis (o-PLSDA) was used as the positively biased control. The projected difference resolution
(PDR) method was used to evaluate the fast GC and fast MS data. FuRES achieved perfect classifications for four models of uncompressed three-way data from two separate runs conducted four days apart. 4
DESI-MS was coupled with single droplet micro-extraction (SDME) to serve as a fast detection method demonstrated by the trace analysis of methamphetamine (MA) in aqueous solution and the detection of an organic reaction product from an ionic liquid (IL). Three-phase liquid SDME was conducted to enrich MA and the subsequent DESI-MS analysis displayed an average enrichment factor of 390-fold for MA. Two-phase liquid SDME was also conducted to directly extract the product of an organic reaction performed in a room temperature IL. The ionization of the resulting droplet extract by DESI allows one to directly examine the reaction product without interference from the ionic liquid.
Panax quinquefolius L samples grown in the United States and in China were quantitatively classified for the first time by using HPLC with an ESI mass spectrometer as the detector combined with chemometric analysis.
Principle component analysis (PCA) and PDR were applied to evaluate the clustering and resolution of classes. With o-PLSDA as the control method, validated by using ten bootstraps and three Latin partitions (BLP), a FuRES classier gave a classification rate of 98 ± 3%, which was equivalent to the classification rate obtained by using o-PLSDA.
Approved: ______
Peter B. Harrington
Professor of Chemistry and Biochemistry 5
Acknowledgments
I would like to thank my research advisor, Dr. Peter de B. Harrington for accepting me as a student. I would not have accomplished my study and finished my dissertation without his invaluable help and support. I would also like to thank my dissertation committee members, Dr. Hao Chen, Dr.
Glen P. Jackson, Dr. Michael Jensen, and Dr. Shigeru Okada for their time and support. Dr. Glen P. Jackson and Dr. Hao Chen are specially thanked for allowing me to use their instruments to conduct my research.
The Department of Chemistry and Biochemistry at Ohio University is thanked for giving me this opportunity to conduct my doctorial research. The
Air Force Research Laboratory in the Wright Patterson Air Force Base and the
Food Composition Laboratory in the U.S. Department of Agriculture are thanked for providing samples for my research projects.
I thank all my group members for their friendships and help. My special thanks go to Weiying Lu and Yao Lu who have been always there for me when I needed help. All my friends are thanked for their encouragement through this journey.
Without the support of my mother Shuzhi Gai, my husband Lixin Ma, my son Xiao Ma, my sister Xiaoguang Sun and my brothers Wenxiang Sun,
Wuxiang Sun, Zhixiang Sun, Daoxiang Sun, I would not have had the courage to pursue a Ph.D degree at this stage of my life. Your support and sacrifice are greatly appreciated. 6
I wish I could tell my father who was such an extraordinary man that it was you who planted the seeds in my heart to pursue my dreams no matter what happens in my life. I will keep working hard to make you proud of me.
7
Table of Contents
Abstract ...... 3
Acknowledgments ...... 5
List of Tables ...... 11
List of Figures ...... 12
List of Abbreviations ...... 14
Chapter 1 Introduction ...... 17
1.1 Fast Gas Chromatography/Fast-Scan Ion trap Mass Spectrometry 18
1.2 Desorption Electrospray Ionization Mass Spectrometry ...... 20
1.3 Single-Droplet Micro-Extraction ...... 26
1.4 Chemometric Techniques ...... 29
1.4.1 Data Preprocessing Techniques ...... 30
1.4.1.1 Wavelet Transformation for Data Size Compression ...... 30
1.4.1.2 Smart Baseline Correction ...... 34
1.4.1.3 Retention Time Alignment ...... 36
1.4.2 Principal Component Analysis ...... 37
1.4.3 Partial Least Squares regression ...... 38
1.4.4 Fuzzy Rule-Building Expert System ...... 41
1.4.5 Projected Difference Resolution ...... 43
Chapter 2 Classification of Jet Fuels by Fuzzy Rule-Building Expert Systems
Applied to Three-Way Data by Fast Gas Chromatography—Fast Scanning
Quadrupole Ion Trap Mass Spectrometry ...... 45
2.1 Introduction ...... 45 8
2.2 Experimental Section ...... 49
2.2.1 Reagents and Sample Preparation ...... 49
2.2.2 Instrumentation and Methods ...... 52
2.2.3 Data Processing ...... 55
2.2.3.1 General Information ...... 55
2.2.3.2 Data Compression ...... 55
2.2.3.3 Principal Component Analysis to Visualize Clustering ...... 55
2.2.3.4 Projected Difference Resolution (PDR) Metric ...... 56
2.2.3.5 FuRES and o-PLSDA Classifiers ...... 56
2.2.3.6 Prediction of a Novel Set of Samples ...... 57
2.3 Results and Discussion ...... 58
2.3.1 Sample Property Analysis by Hierarchical Cluster Analysis and
PCA 58
2.3.2 Instrumentation and GC/MS Data ...... 61
2.3.2.1 Data Compression ...... 64
2.3.2.2 PCA Results ...... 65
2.3.2.3 PDR Results ...... 67
2.3.2.4 FuRES and o-PLSDA Classification Results ...... 67
2.3.2.5 FuRES Model ...... 74
2.3.2.6 Prediction of a Novel Collection of Samples ...... 76
2.4 Conclusions ...... 79
Chapter 3 Coupling of Single Droplet Micro-Extraction with Desorption
Electrospray Ionization-Mass Spectrometry ...... 81 9
3.1 Introduction ...... 81
3.2 Experimental Section ...... 83
3.2.1 Chemicals and Reagents ...... 83
3.2.2 Three-phase SDME for Trace Analysis of MA ...... 84
3.2.2.1 Sample Preparation for MA Extraction ...... 84
3.2.2.2 Extraction Procedure for MA ...... 84
3.2.2.3 Calibration of DESI-MS Detection of MA ...... 86
3.2.3 Detection of the Products of Organic Reactions Performed in
Ionic Liquids ...... 87
3.2.4 Instrumentation ...... 88
3.3 Results and Discussion ...... 89
3.3.1 SDME/DESI-MS for Trace Analysis of MA ...... 89
3.3.1.1 Optimization of Three-Phase SDME for MA Extraction ...... 89
3.3.1.2 Optimization of the DESI Apparatus for MA Detection ...... 90
3.3.1.3 Calibration of Single Droplet DESI-MS Analysis for MA ...... 91
3.3.1.4 Extraction Result and Enrichment Factor for MA ...... 94
3.3.2 SDMS/DESI-MS Detection of Products in Organic Reactions with
ILs as Solvents ...... 96
3.3.3 High Throughput Analysis ...... 100
3.4 Conclusions ...... 100
Chapter 4 Classification of Cultivation Locations of Panax Quinquefolius L
Samples by High Performance Liquid Chromatography—Electrospray
Ionization Mass Spectrometry and Chemometric Analysis ...... 102 10
4.1 Introduction ...... 102
4.2 Experimental Section ...... 105
4.2.1 Samples, Reagents and Sample Pretreatment ...... 105
4.2.2 Instrumentation ...... 108
4.2.3 HPLC—MS Data Collection Conditions ...... 108
4.2.4 Data Pretreatment ...... 109
4.2.4.1 Baseline Correction ...... 109
4.2.4.2 RT Alignment ...... 109
4.2.5 Classification by the FuRES Classifier with Optimized-PLS as the
Control Method ...... 110
4.3 Results and Discussion ...... 110
4.3.1 Optimization of Data Collection Methods ...... 110
4.3.2 Pretreatment Effects of Data ...... 112
4.3.3 PCA Results ...... 116
4.3.4 PDR Results ...... 117
4.3.5 FuRES and o-PLSDA Classification Results ...... 118
4.3.6 FuRES Model ...... 121
4.4 Conclusions ...... 131
Chapter 5 Summary and Future Work ...... 133
References...... 137
Appendix A: Publications ...... 163
Appendix B: Presentations ...... 164
11
List of Tables
Page
Table 2-1. Types, IDs and available properties of the samples used in the comparison of the FF-CI mode with the FN-CI mode ...... 50
Table 2-2. Types, IDs and available properties of the samples used in the prediction of unknown samples under the FF-CI mode ...... 51
Table 2-3. GC separation and MS scan conditions for the FF-CI mode ...... 54
Table 2-4. Prediction and resolution results of FF-CI and FN-CI with and without wavelet compression ...... 69
Table 2-5. Confusion matrix of FuRES classification under the FF-CI mode with wavelet compression ...... 73
Table 2-6. Comparison of prediction errors for data collected four days apart with and without wavelet compression and RT alignment...... 77
Table 4-1. Ginseng samples grown in two countries used in this work. AM and CH denote the samples grown in the United States and China, respectively ...... 107
Table 4-2. Table 1. Classification rates with 95% confidence intervals by FuRES and o-PLSDA classifiers with no alignment (-A) and RT aligned (+A) and without baseline correction (-B) and with baseline correction (+B). The asterisks (*) denote alignments that were applied to the calibration set and the prediction data were fit to the mean of the aligned training data...... 119
Table 4-3. Ten largest magnitude peaks in the two-way FuRES rule ...... 123 12
List of Figures
Page
Figure 1-1. Schematic representation of solid desorption electrospray ionization mass spectrometry...... 22
Figure 1-2. Schematic representation of liquid DESI...... 25
Figure 1-3. Schematic representation of a three-phase single-droplet micro- extraction set-up...... 27
Figure 1-4. A flow chart illustrating the pyramid algorithm used for WT data compression...... 33
Figure 1-5. Graphical illustration of PLS regression...... 40
Figure 2-1. Dendrogram of selected properties of jet fuel samples 3528, .. 59
Figure 2-2. PCA score plotting of selected properties of samples 3528, ..... 60
Figure 2-3. Reconstructed TIC chromatograms (A) and average mass ...... 62
Figure 2-4. Two-way data image of sample 2885 under the FF-CI mode ... 64
Figure 2-5. The principal component score plotting of data collected with .. 66
Figure 2-6. FuRES classification tree of jet fuels of data collected under the FF-CI mode...... 75
Figure 3-1. Schematics showing (A) the three-phase SDME set-up for MA analysis and (B) the two-phase SDME set-up for DPTU detection from an ionic liquid; (C) image showing the direct desorption and ionization of the obtained single droplet extract contained in the syringe with a needle of the deactivated fused silica capillary...... 86
Figure 3-2. Representative DESI-MS spectrum of an 800. ng/mL MA standard solution for calibration. Inset: standard calibration line for the DESI-MS detection of single droplet extract samples. The ratios of peak heights of m/z 150 to those of m/z 74 were plotted versus the concentrations of standard MA solutions. (-) represents the upper and lower boundaries of 13 the 95% confidence intervals; (—) represents the calibration line obtained from the average data points of three replicates of standard solutions...... 93
Figure 3-3. Colision induced dissociation mass spectrum of m/z 150...... 94
Figure 3-4. DESI-MS spectra of three extraction replicates (a single scan is displayed for each spectrum of three replicates A–C)...... 95
Figure 3-5. (A) DESI-MS spectrum of the extracted DPTU from [Bmim][BF4] by using CCl4 as the extractant; (B) CID MS/MS mass spectrum of the protonated reaction product [DPTU+H]+ (m/z 229)...... 99
Figure 4-1. Two-way image (log values) of sample WSQ2 from one HPLC—MS run...... 111
Figure 4-2. A demonstration of baseline correction effect...... 113
Figure 4-3. Reconstructed total ion current chromatograms with baseline 115
Figure 4-4. PCA score plots of samples before and after RT alignment with ...... 117
Figure 4-5. FuRES classification tree with the two-way 3-dimensional ..... 122
Figure 4-6. The extracted ion chromatogram (A) of ion m/z 574.3 and the corresponding mass spectra of the major peaks (B)...... 125
Figure 4-7. PCA score plot with only first two largest magnitude peaks in the two-way FuRES rule...... 128
Figure 4-8. Mass spectra of all 12 P. quinquefolius L samples grown in the United States (A) and all 12 P. quinquefolius L samples grown in China (B)...... 130
14
List of Abbreviations
ANN ...... artificial neural network
ANOVA ...... analysis of variance
APCI ...... atmospheric-pressure chemical ionization
CE ...... capillary electrophoresis
CI ...... chemical ionization
CID ...... collision-induced dissociation
DESI ...... desorption electrospray ionization
DPTU ...... N,N’-diphenylthiourea
ECD ...... electron capture detector
ESI ...... electrospray ionization
ESSI ...... electrosonic spray ionization
FID ...... flame-ionization detector
FT ...... Fourier transformation
FuRES ...... fuzzy rule-building expert system
GC ...... gas chromatography
HPLC...... high performance liquid chromatography
IL ...... ionic liquid
IR ...... infrared
KNN ...... K-Nearest neighbor
LDA ...... linear discriminant analysis
LOD ...... limit of detection
LOQ ...... limit of quantification 15
MA ...... methamphetamine
MALDI ...... matrix-assisted laser desorption/ionization
MLS ...... multivariate least squares
MLR ...... multiple linear regression
MS ...... mass spectrometry
MSD ...... mass spectrometer detector
MRE ...... mass range extension
NIR ...... near infrared
NRL ...... the Naval Research Lab
ODS ...... octadecylsilane o-PLSDA ...... optimized-partial least squares discriminant analysis
PC ...... principal component
PCA ...... principal component analysis
PCT ...... principal component transform
PDR ...... projected difference resolution
PLS ...... partial least squares
P. quinquefolius L ...... Panax quinquefolius L
PRESS ...... predicted residual error sum of squares
PVDF ...... polyvinylidene fluoride
QIT ...... quadrupole ion trap
RT ...... retention time
RTIL ...... room temperature ionic liquid
SDME ...... single droplet micro-extraction 16
SIMCA ...... soft independent modeling class analogy
SPE ...... solid phase extraction
SPME ...... solid phase micro-extraction
SVD ...... singular value decomposition
TCD ...... thermal conductivity detector
TIC ...... total ion chromatogram
TLC ...... thin layer chromatography
ToF ...... time-of-flight
WT ...... wavelet transformation
17
Chapter 1 Introduction
Forensic applications of instrumental analysis techniques combined with chemometric multivariate data analysis are presented in this dissertation. Chapter 1 presents general introduction to instrumental analysis techniques and chemometric data treatment methods used in this dissertation. Chapter 2 presents the forensic application of fast gas chromatography (GC)/fast scan mass spectrometry (MS) for the classification of jet fuel samples by their lot numbers. The forensic application of single- droplet micro-extraction (SDME) coupled with desorption electrospray ionization-mass spectrometry (DESI-MS) for the detection of illicit drugs is focused on in Chapter 3. The extended application of SDME coupled with
DESI-MS for the detection of reaction products in room temperature ionic liquids is also presented in Chapter 3. Chapter 4 presents the classification of Panax quinquefolius L (P. quinquefolius L) samples grown in different geographical locations (United States and China ) by using two-way datasets collected by using high performance liquid chromatography—electrospray mass spectrometry (HPLC—MS) combined with chemometric analysis.
Chapter 5 introduces the summary and future work. Appendices A and B are respectively selected publications and presentations associated with this dissertation. 18
1.1 Fast Gas Chromatography/Fast-Scan Ion trap Mass
Spectrometry
Since the first development in the 1950s, GC has been a significant separation method used for separation and analysis of volatile compounds.1
In analytical chemistry, it is always desirable to spend less time on analysis while sufficient information can be obtained. Fast GC separation can shorten analysis time and increase throughput. According to the definition given by
Matisová et al., fast GC has a separation time per sample of within a few minutes and a speed enhancement factor of 5–30, and a peak width at half- height of 1-3 s while the plate number can be kept comparable to the conventional GC.2 Compared with conventional GC, fast GC offers the advantages of a tremendous improvement in laboratory throughput, a much lower cost per sample, a shorter time needed for the analysis, and especially the possibility of the usage as an online monitoring means. Strategies toward a fast-GC are presented as follows.
The first strategy to implement fast-GC is to change some hardware conditions while a constant resolution is preserved, such as reducing the column diameter, selecting a low molecular mass gas as the mobile phase, and/or applying a vacuum outlet for the separation system.
Secondly, only a resolution necessary for the separation of the targeted analytes should be obtained. Although only a resolution greater than 1.5 is considered baseline-resolved in theory, a resolution much less than 1 may be sufficient for some cases. To achieve an acceptable resolution 19 and a fast analysis, a short column with a small inner diameter can be used.
In addition, a flow rate greater than the optimum flow rate specified by the van Deetmer equation, higher temperatures and a faster temperature ramp rate in the temperature profile, and a less column coating thickness can be selected.
Fast GC suffers from limited chromatographic resolution especially when it is used to separate very complicated samples such as jet fuels. It also requires a fast detection method to match the fast separation speed.
Commonly used detectors for GC are the electron capture detector (ECD), the flame ionization detector (FID), the thermal conductivity detector (TCD), and the mass spectrometric detector (MSD). Among them, the MSD is one of the most common, sensitive, and informative detectors because the MSD can provide information of the molecular structures of the compounds. Time-of- flight (ToF) mass spectrometers are capable of very fast data acquisition rates (kHz duty cycles are possible.) as summarized in a review GC/ToF-MS application.3 However, ToF mass spectrometers have disadvantages of relatively higher costs to purchase and maintain as well as more stringent vacuum requirements than ion trap mass spectrometers.
Most conventional scanning MS methods, such as quadrupole ion trap
(QIT) MS, have a limited scan rate of 5500 Th/s (i.e., a scan rate parameter of 0.18 ms/Th), or a spectrum acquisition rate of about 3 Hz. This scan rate will be insufficient when it is used as a detection method for fast separation methods such as fast GC. Here the scan rate with a unit of (m/z)/s or Th/s is 20 the rate the ions are scanned out of the ion trap. Yang and Bier proposed a fast ion trap MS scan strategy with a scan rate as fast as 66 660 Th/s (i.e., a scan rate parameter of 0.015 ms/Th), which is 12 times greater than the scan rate of conventional-QIT-MS.4 Yang and Bier noted a fortuitous and unique result of increasing the scan rate of the QIT, an overall signal to noise improvement (an increase of n1/2 in which n is the number of scans conducted over the period for a normal scan). The improvement is achieved through the reduction in space charge effects and a decrease in peak widths
(hence an increase in peak heights).4
The primary advantage of fast scanning in QITs is that it is better suited for coupling with time-limited fast chromatographic separations that yield peaks in narrow time windows, enhancement in signal to noise ratio and improvement in limit of detection (LOD) of trace analytes. Fast scanning ion traps can yield much more information than their conventional scanning counterparts. Pattern recognition methods can exploit the extra information afforded from the faster scanning instruments to overcome difficulties such as limited chromatographic resolution. In Chapter 2, a project utilizing fast-
GC/fast-scan QIT-MS as three-way data collection method is described in details.
1.2 Desorption Electrospray Ionization Mass Spectrometry
Desorption electrospray ionization mass spectrometry (DESI-MS) was first introduced by Cooks and coworkers in 2004, which is an ionization technique conducted in open air under ambient conditions based on 21 electrospray ionization mass spectrometry (ESI-MS).5, 6 DESI can be used in mass spectrometry for analysis of liquids and solids. DESI has been used for a huge variety of analytes encompassing illicit drugs,7, 8 chemical warfare agents,9-13 pharmaceuticals,14-19 biological samples,20-23, etc.24, 25 DESI-MS is similar to ESI-MS by operating under ambient conditions and capable of analyzing biomolecules with high molecular masses. DESI-MS offers the advantages of no or little sample pretreatment, short analysis time (as short as a few seconds), equivalent specificity to other mass spectrometric methods, high throughput and versatility. In Figure 1-1, a typical DESI source consists of an electrosonic spray ionization (ESSI) solvent tubing in which a solvent is applied with a high voltage, a nebulizing gas which is usually nitrogen and a surface as the analyzing platform for samples. 22
Figure 1-1. Schematic representation of solid desorption electrospray ionization mass spectrometry.
Adapted from http://www.aocs.org/Membership/FreeCover.cfm?itemnumber=1108
(Accessed April 26)
23
Two mechanisms have been proposed for DESI depending on the types of analytes.25 The first mechanism is the droplet pick-up mechanism.
When analytes such as proteins and peptides that are usually analyzed by
ESI-MS are sampled by DESI, the mechanism is similar to that of ESI in the formation of charge droplets containing charged analyte species.25 Firstly, the ESSI droplets are ejected onto the surface where the analyte is placed and spread out on the surface. At the edge of the droplet spot charged droplets containing analyte species are produced and driven into the MS inlet by the gas in the spray, Coulombic effects and vacuum of the MS inlet.
Under this mechanism, the ionization efficiency is highly related to the velocity of the ESSI spray, which is determined by both the gas flow rate and the solution flow rate, and the distance between the spray and the surface.
This mechanism is supported by the fact that when an equally high voltage is applied to the surface where analytes are placed, ionization still occurs.11
The second mechanism is believed to be a charge transfer process between the ESSI solvent ions and the analyte molecules that are suitable for atmospheric-pressure chemical ionization (APCI) analysis. The proof of this mechanism is that the ionization occurs only when a large difference
(greater than 2 kV) occurs between the voltage applied to the ESSI solvent and that of the surface. It is proposed that heterolytic charge-transfer reactions including the proton exchange electron exchange between gas- phase ions and surface molecules are responsible for the detection of analyte ions in the MS as illustrated in Equation (1-1). 24