COMPARISON OF GAS CHROMATOGRAPHY-MASS SPECTROMETRY AND LIQUID
CHROMATOGRAPHY-MASS SPECTROMETRY FOR DISCRIMINATION OF SALVIA
DIVINORUM FROM RELATED SALVIA SPECIES USING CHEMOMETRIC
PROCEDURES
By
Monica L. Bugeja

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Forensic Science
2012

 
 
 

ABSTRACT
COMPARISON OF GAS CHROMATOGRAPHY-MASS SPECTROMETRY AND LIQUID
CHROMATOGRAPHY-MASS SPECTROMETRY FOR DISCRIMINATION OF SALVIA
DIVINORUM FROM RELATED SALVIA SPECIES USING CHEMOMETRIC
PROCEDURES
By
Monica L. Bugeja

Salvia divinorum is a hallucinogenic plant that has recently gained legislative attention
due to an increase in its recreational use. S. divinorum and/or its active constituent, salvinorin A,
are currently regulated in 26 states, including Michigan. As there are over 900 known Salvia
species, discrimination of S. divinorum from other Salvia species is important in a forensic
laboratory. Salvinorin A is known to be found only in S. divinorum thus far. Therefore, the
presence of salvinorin A, detected using gas chromatography-mass spectrometry (GC-MS),
serves as the current method for identifying S. divinorum. However, salvinorin A can be
extracted from S. divinorum to prepare liquid extracts for recreational use. This can result in low
levels of salvinorin A in the residual plant material that may or may not be detectable, causing
possible difficulty with identification. Hence, an alternative procedure for identifying S.
divinorum, such as using a full chemical fingerprint, would be beneficial.
The purpose of this research was to generate chemical fingerprints of S. divinorum and
four related Salvia species using GC-MS to analyze the volatile compounds and liquid
chromatography-mass spectrometry (LC-MS) to analyze the non-volatile compounds. The
differentiation of S. divinorum from the other species based on chemical fingerprints was then
investigated using statistical procedures.

ACKNOWLEDGMENTS

The completion of this thesis could not be possible without the help and
encouragement of several individuals. First, I would like to thank my family for listening to
me complain and stress throughout my years of schooling, supporting me financially as I
continued my education, and raising me the way I am. I would not be the person I am today
without everything my family has taught me. In addition, I would like to thank a good friend
and boyfriend, Daniel Kelly, who has given me advice, support, and encouragement
throughout this project. I would also like to thank him for the time he has spent coming to
visit me and taking me away from this thesis project every once in a while to be able to laugh
and relax. Furthermore, I would like to thank the Forensic Chemistry Group, who have
supported and assisted me with this research. I would especially like to recognize John
McIlroy, who has taught me so much throughout the year. I am so grateful for all of his help
and could not thank him enough. I would like to recognize Melissa Bodnar Willard, whose
previous research on S. divinorum created a path to this research. I would also thank her for
all of her guidance. I would like to thank Suzanne Towner, who has been a wonderful,
supportive, encouraging, and helpful friend throughout this project, and Tiffany VanDeMark,
who has also been a supportive and kind friend. I would like to acknowledge my friends and
fellow scientists at the Michigan State Police, who have continuously supported and
motivated me throughout my schooling.
iii

In addition, I would like to thank the Midwest Forensics Resource Center for funding
this project. Dr. Daniel Jones and staff at Michigan State University’s Mass Spectrometry
Facility for access to and help with the liquid chromatograph-mass spectrometer and
interpretation of mass spectra. I would also like to thank Dr. Kevin Walker for access to the
Salvia plants and Jim Klug and staff at Michigan State University’s research greenhouse for
maintaining the Salvia plants throughout this project. Finally, I would like to thank my
committee members, Dr. Ruth Waddell Smith, Dr. Victoria McGuffin, and Dr. Mahesh Nalla
for their help, guidance, and patience.

iv

 
 

TABLE OF CONTENTS

List of Tables

vii

List of Figures

x

Chapter 1: Introduction
1.1 What is Salvia divinorum?
1.2 Use, Abuse, and Effects of Salvia divinorum
1.3 Regulation
1.4 Extraction and Analysis of Salvia divinorum
1.4.1 Extraction Procedures
1.4.2 Analysis Procedures
1.4.3 Statistical Procedures
1.5 Identification of other Salvia species
1.5.1 S. guaranitica
1.5.2 S. nemorosa
1.5.3 S. officinalis
1.5.4 S. splendens
1.6 Objective
REFERENCES

1
1
3
4
4
4
6
8
11
11
11
12
13
13
15

Chapter 2: Theory
2.1 Instrumental Techniques
2.1.1 Gas Chromatography-Mass Spectrometry
2.1.2 Liquid Chromatography-Mass Spectrometry
2.2 Data Pretreatment
2.2.1 Background Correction
2.2.1 Smoothing
2.2.3 Retention Time Alignment
2.2.4 Normalization
2.3 Data Analysis
2.3.1 Pearson Product Moment Correlation Coefficients
2.3.2 Principal Components Analysis
2.3.3 Student’s t-tests
REFERENCES

19
19
19
31
40
41
42
43
45
45
45
46
48
50

Chapter 3: Association and Discrimination of Salvia divinorum from Related Salvia
Species Based on Chemical Fingerprints Generated Using Gas ChromatographyMass Spectrometry
3.1 Introduction

54
54

v
 

3.2 Material and Methods
3.2.1 Extraction of Salvia Species
3.2.2 Gas Chromatography-Mass Spectrometry Analysis
3.2.3 Data Pretreatment
3.2.4 Data Analysis
3.3 Results and Discussion
3.3.1 Provisional Identification of Compounds Present in Each Species
3.3.2 Investigation of Extraction and Analysis Precision Using Pearson
Product Moment Correlation Coefficients
3.3 Association and Discrimination of S. divinorum from Other Salvia Species
3.4 Summary
REFERENCES

54
54
57
58
59
60
60
66
69
85
86

Chapter 4: Association and Discrimination of Salvia divinorum from Related Salvia
Species Based on Chemical Fingerprints Generated Using Liquid ChromatographyMass Spectrometry
88
4.1 Introduction
88
4.2 Material and Methods
89
4.2.1 Investigation of Extraction Solvent
89
4.2.2 Investigation of Liquid Chromatography-Mass Spectrometry
Ionization Mode
90
4.2.3 Extraction of Salvia Species
91
4.2.4 Liquid Chromatography-Mass Spectrometry Analysis
91
4.2.5 Data Pretreatment
92
4.2.6 Data Analysis
93
4.3 Results and Discussion
94
4.3.1 Investigation of Extraction Solvent and Ionization Mode
94
4.3.2 Provisional Identification of Compounds Present in Each Species
94
4.3.3 Investigation of Extraction and Analysis Precision Using Pearson
Product Moment Correlation Coefficients
101
4.3.4 Association and Discrimination of Salvia divinorum from Other Salvia
Species
105
4.4 Summary
121
REFERENCES
123
Chapter 5: Conclusions and Future Work
5.1 Gas Chromatography-Mass Spectrometry Analysis of Salvia divinorum and
Related Salvia Species
5.2 Liquid Chromatography-Mass Spectrometry Analysis of Salvia divinorum and
Related Salvia Species
5.3 Comparison of Gas Chromatography-Mass Spectrometry and Liquid
Chromatography-Mass Spectrometry for the Analysis of Salvia divinorum and
Related Salvia Species
5.4 Utility of Gas Chromatography-Mass Spectrometry and Liquid
Chromatography-Mass Spectrometry for the Analysis of Salvia divinorum
5.5 Future Work
vi
 

126
126
127

127
128
129

REFERENCES

131

Appendices
133
Appendix A: Provisional Identification of Compounds in the Five Salvia Species
Analyzed Using Gas Chromatography-Mass Spectrometry
134
Appendix B: Provisional Identification of Compounds in the Five Salvia Species
Analyzed Using Liquid Chromatography-Mass Spectrometry
141

 

vii
 

LIST OF TABLES

Table 3.6: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients for replicate injections of the same extract

67

Table 3.7: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients (n=36) for all extracts for each species

68

Table 3.8: Absolute value of tcalculated and tcritical values calculated using two-tailed
Student’s t-tests performed on the mean scores of principal component 1 (PC1) and
principal component 2 (PC2) for comparisons (n=18) of S. divinorum to the other
Salvia species*

80

Table 3.9: Absolute value of tcalculated and tcritical values calculated using two-tailed
Student’s t-tests performed on the mean scores of principal component 1 (PC1) and
principal component 2 (PC2) for comparisons (n=18) between S. guaranitica,
S. nemorosa, and S. splendens

81

Table 3.10: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients for comparisons (n=81) between Salvia species

83

Table 4.3: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients for replicate injections of the same extract

102

Table 4.4: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients for all extracts for each species

104

viii
 

Table 4.7: Absolute value of tcalculated and tcritical values calculated using two-tailed
Student’s t-tests performed on the mean scores of principal component 1 (PC1) and
principal component 2 (PC2) for comparisons (n=18) of S. divinorum to the other
Salvia species*

116

Table 4.8: Absolute value of tcalculated and tcritical values calculated from Student’s t-tests
performed on the mean scores of principal component 1 (PC1) and principal
component 2 (PC2) between S. guaranitica, S. nemorosa, and S. splendens (n=18)*
117
Table 4.9: Mean, standard deviation, and range of Pearson product moment
correlation (PPMC) coefficients for comparisons (n=81) between Salvia species

ix
 

119

 
 
 

LIST OF FIGURES

Figure 1.1: Chemical structure of salvinorin A.

2

Figure 2.1: Schematic diagram of a gas chromatograph.

21

Figure 2.2: Schematic diagram of an electron ionization source adapted from [1].

25

Figure 2.3: Schematic diagram of a quadrupole mass analyzer adapted from [1, 2].

27

Figure 2.4: Schematic diagram of a continuous dynode electron multiplier
adapted from [1].

28

Figure 2.5: Mass spectrum of salvinorin A analyzed using electron ionization.

30

Figure 2.6: Schematic diagram of a liquid chromatograph adapted from [5].

32

Figure 2.7: Schematic diagram of electrospray ionization source operating in positive
ionization mode adapted from [1].

35

Figure 2.8: Schematic diagram of a hybrid triple quadrupole/linear ion trap
adapted from [1].

37

Figure 2.9: Mass spectrum of salvinorin A analyzed using electrospray ionization.

39

Figure 3.1: Schematic diagram of the filtration system.

56

x
 

Figure 3.2: Total ion chromatogram of Salvia divinorum extracted in dichloromethane
and analyzed using gas chromatography-mass spectrometry. Provisional identifications
based on library searches and literature sources are as follows: 1=heptacosane, 2=squalene,
3=progesterone, 4=nonacosane, 5=an acetate, 6=hentriacontane, 7=salvinorin B,
8=salvinorin A, 9=tetratriacontane, 10=salvinorin D, 11=sitosterol, and
12=pentatriacontane [1, 2]. See Appendix A, Table A.1 for more detailed information.

61

Figure 3.3: Total ion chromatogram of Salvia guaranitica extracted in dichloromethane
and analyzed using gas chromatography-mass spectrometry. Provisional identifications
based on library searches and literature sources are as follows: 2= squalene,
3=progesterone, 6= hentriacontane, 9=tetratriacontane, 11=sitosterol, 13=phthalide,
14=β-cubebene, 15= an ester, 16= an alkene, 17=vitamin E, 18=dione, 19=dione,
20=nitrolepidine, 21=benzenesulfonamide, 22=β-amyrin, and 23=α-amyrin [5, 6]. See
Appendix A, Table A.2 for more detailed information.

62

Figure 3.4: Total ion chromatogram of Salvia nemorosa extracted in dichloromethane
and analyzed using gas chromatography-mass spectrometry. Provisional identifications
based on library searches and literature sources are as follows: 3=progesterone,
4=nonacosane, 6=hentriacontane, 9=tetratriacontane, 11=sitosterol, 14=β-cubebene,
17=vitamin E, 24=β-caryophyllene, 25=alkene, 26=amide, 27=phenyl,
28=11-n-decyltetracosane, and 29=hextriacontane [4]. See Appendix A, Table A.3 for
more detailed information.

63

Figure 3.5: Total ion chromatogram of Salvia officinalis extracted in dichloromethane
and analyzed using gas chromatography-mass spectrometry. Provisional identifications
based on library searches and literature sources are as follows: 1=heptacosane,
2=squalene, 3=progesterone, 4=nonacosane, 6=hentriacontane, 9=tetratriacontane,
11=sitosterol, 17=vitamin E, 23=α-amyrin, 25=alkene, 29=hexatriacontane,
30=α-pinene, 31=camphene, 32=β-pinene, 33= β-myrcene, 34=eucalyptol, 35=thujone,
36=thujone, 37=camphor, 38=borneol, 39=bornyl acetate and/or isoborynyl acetate,
40=bicyclic compound, 41=a cyclopropane compound, 42=α-caryophyllene,
43=epiglobulol, 44=diene, 45=epimanool, 46=isocarnosol, 47=phenyl, 48=dione,
49=squalane, 50=triacontane, 51=dotriacontane, 52=9-n-octylhexacosane, and
53=oleanolic acid [3]. See Appendix A, Table A.4 for more detailed information.

64

Figure 3.6: Total ion chromatogram of Salvia splendens extracted in dichloromethane
and analyzed using gas chromatography-mass spectrometry. Provisional identifications
based on library searches and literature sources are as follows: 2=squalene,
3=progesterone, 4=nonacosane, 6=hentriacontane, 9=tetratriacontane, 11=sitosterol,
17=vitamin E, 23=α-amyrin, 25=alkene, 29=hextriacontane, 53=oleanolic acid,
54=8-episalviarin, and 55=cardenolide [7]. See Appendix A, Table A.5 for more
detailed information.
Figure 3.8: Scores plot of principal component 1 versus principal component 2
showing discrimination of S. divinorum from the other four Salvia species based on
chemical profiles obtained by gas chromatography-mass spectrometry.
xi
 

65

70

Figure 3.9: Loading plots for (A) principal component 1 (PC1) and (B) principal
component 2 (PC2) based on principal components analysis of the five Salvia species
analyzed by gas chromatography-mass spectrometry. Selected compounds are labeled
based on species as follows: a=Salvia divinorum, b=Salvia guaranitica, c=Salvia
nemorosa, d=Salvia officinialis, e=Salvia splendens. IS=internal standard. See
Appendix A for provisional identifications based on library searches and
literature sources.

71

Figure 3.10: Magnified version of nine total ion chromatograms of S. officinalis
(solid line) and one total ion chromatogram of S. nemorosa (dashed line) showing
the misalignment of the internal standard peak.

73

Figure 3.11: (A) Mean-centered data of one replicate of S. officinalis and (B) loadings
for the same replicate of S. officinalis on principal component 2 based on chemical
profiles obtained by gas chromatography-mass spectrometry with progesterone added
as the internal standard (IS).

74

Figure 3.12: (A) Mean-centered total ion chromatogram of one replicate of S. nemorosa
and (B) loadings for the same replicate of S. nemorosa on principal component 1 based
on chemical profiles obtained by gas chromatography-mass spectrometry with
35=thujone, 37=camphor, and 45=epimanool.

77

Figure 4.1: Total ion chromatograms of S. divinorum extracted in (A) acetonitrile
with progesterone added as an internal standard (IS) and (B) acetonitrile/2-propanol/water.
Both extracts were analyzed using positive ionization mode. Provisional identifications
based on literature sources, are as follows: 1=salvinorin B, 2=salvinorin D,
3=salvinorin F, 4=unidentified compound A, 5=unidentified compound B,
6=salvinorin A, 7=salvinorin C, 8=progesterone (IS), 9=unidentified compound C,
10=unidentified compound D, and 11=unidentified compound E [2, 3, 6, 7].
95
Figure 4.2: Total ion chromatogram of Salvia divinorum extracted in acetonitrile
and analyzed in positive mode using liquid chromatography-mass spectrometry.
Provisional identifications based on literature sources are as follows: 1=salvinorin B,
2=terpenoid (molecular mass= m/z 406), 3=terpenoid (molecular mass= m/z 378),
4=salvinorin D, 5=salvinorin F, 6=salvinorin A, 7=salvinorin C, 8=sesquiterpenoid,
and 9=unidentified compound (molecular mass= m/z 426) [2, 3, 6]. See Appendix B,
Table B.1, for more detailed information.
Figure 4.3: Total ion chromatogram of Salvia guaranitica extracted in acetonitrile
and analyzed in positive mode using liquid chromatography-mass spectrometry.
Provisional identifications based on literature sources are as follows: 8=sesquiterpenoid,
10=diterpenoid (molecular mass=m/z 340), 11=terpenoid (molecular mass=m/z 398),

xii
 

96

12= terpenoid (molecular masses=m/z 340, 398), 13=terpenoid (molecular mass=m/z 398),
14=oleanolic/ursolic acid, 15=vitamin E, and 16=unidentified compound [8–12]. See
Appendix B, Table B.2, for more detailed information.
97
Figure 4.4: Total ion chromatogram of Salvia nemorosa extracted in acetonitrile
and analyzed in positive mode using liquid chromatography-mass spectrometry.
Provisional identifications based on literature sources are as follows: 8=sesquiterpenoid,
14=oleanolic/ursolic acid, 15=vitamin E, and 17=unidentified compound [8–12]. See
Appendix B, Table B.3, for more detailed information.

98

Figure 4.5: Total ion chromatogram of Salvia officinalis extracted in acetonitrile
and analyzed in positive mode using liquid chromatography-mass spectrometry.
Provisional identifications based on literature sources are as follows: 8=sesquiterpenoid,
14=oleanolic/ursolic acid, 15=vitamin E, 18=3-O-methylkaempferol,
19=carnosol and carnosic acid, 20=methyl carnosate/rosmanol/ epirosmanol, and
21=unidentified compound [8–12, 13–17]. See Appendix B, Table B.4, for more detailed
information.

99

Figure 4.6: Total ion chromatogram of Salvia splendens extracted in acetonitrile
and analyzed in positive mode using liquid chromatography-mass spectrometry.
Provisional identifications based on literature sources are as follows: 8=sesquiterpenoid,
9=unidentified compound (molecular mass=m/z 426),14=oleanolic/ursolic acid,
15=vitamin E, 22=diterpenoid (molecular mass=m/z 400), 23=diterpenoid
(molecular mass=m/z 342), 24=diterpenoid (molecular mass=m/z 362), and
25=phenolic methyl ether and an unidentified compound (molecular masses=m/z 406,
548) [8–12]. See Appendix B, Table B.5, for more detailed information.

100

Figure 4.7: Scores plot of principal component 1 versus principal component 2
showing discrimination of S. divinorum from the other four Salvia species based on
chemical profiles obtained by liquid chromatography-mass spectrometry.

106

Figure 4.8: Loadings plots for (A) principal component 1 (PC1) and
(B) principal component 2 (PC2) based on principal components analysis of the
five Salvia species analyzed by liquid chromatography-mass spectrometry with
1=salvinorin B, 2=terpenoid , 3=terpenoid, 4=salvinorin D, 5=salvinorin F,
6=salvinorin A, 7=salvinorin C, 8=sesquiterpenoid, 9=unidentified compound,
10=diterpenoid, 11=terpenoid, 12=terpenoid, 13=terpenoid, 14=oleanolic/ursolic acid,
15=vitamin E, 16=unidentified compound, 17=unidentified compound, 18=3-Omethylkaempferol, 19=carnosol and carnosic acid, 20=methyl carnosate/ rosmanol/
epirosmanol, 21=unidentified compound, 22=diterpenoid, 23=diterpenoid,
24=diterpenoid, 25=phenolic methyl ether and unidentified compound.

107

Figure 4.9: (A) Mean-centered total ion chromatogram of S. divinorum and
(B) loadings for S. divinorum on principal component 1 based on chemical profiles
obtained by liquid chromatography-mass spectrometry with
19=carnosol and carnosic acid, and 20=methyl carnosate/rosmanol/epirosmanol.

109

xiii
 

Figure 4.10: (A) Mean-centered total ion chromatogram of S. nemorosa and
(B) loadings for S. nemorosa on principal component 2 based on chemical profiles
obtained by liquid chromatography-mass spectrometry with 6=salvinorin A,
9=unidentified compound, 13=terpenoid, 15=vitamin E, 16=unidentified compound,
17=unidentified compound, 19=carnosol and carnosic acid, 20=methyl
carnosate/rosmanol/epirosmanol, and 21=unidentified compound.

111

Figure 4.11: (A) Mean-centered total ion chromatogram of S. splendens, and
(B) loadings for S. splendens on principal component 2 based on chemical profiles
obtained by liquid chromatography-mass spectrometry with 6=salvinorin A,
9=unidentified compound, 13=terpenoid, 15=vitamin E, 16=unidentified compound,
17=unidentified compound, 21=unidentified compound, and
25=phenolic methyl ether and unidentified compound.

113

 

xiv
 

Chapter 1: Introduction
1.1 What is Salvia divinorum?
Salvia divinorum is a hallucinogenic plant originating from Oaxaca, Mexico that has
recently gained legislative attention due to an increase in its recreational use [1-8]. There are
over 900 Salvia species that are divided into three clades based on genetics and plant structure,
as well as geographical origin [9]. Of these three clades, S. divinorum belongs to clade II and is a
member of the Lamiaceae family [6].
S. divinorum consists of several chemical compounds; however, only one chemical
compound is known to contribute to its hallucinogenic nature, a trans-neoclerodane diterpene
named salvinorin A [1, 2, 4, 5]. This compound was first isolated from S. divinorum by Ortega et
al. in 1982 [5, 10]. Ortega et al. determined the structure of salvinorin A (Figure 1.1) using
13

infrared (IR) spectroscopy, carbon and hydrogen nuclear magnetic resonance ( CNMR and
1

HNMR, respectively), single-crystal X-ray analysis, and mass spectrometry (MS) [10]. Using

these instrumental techniques, the molecular formula of salvinorin A was determined to be
C23H28O8. Using MS with electron ionization, salvinorin A had a molecular ion of m/z 432, and
a base peak of m/z 94 [3].
Other chemical compounds identified in S. divinorum include salvinorins A through I,
divinatorins A through F, salvidivins A through D, and hardwickiic acid [1, 2, 6, 11, 12].
Salvinorin A is typically the most abundant compound, followed by salvinorins B, C, and D.
While salvinorins E and F, divinatorins A through C, and hardwickiic acid may be present, these
compounds are typically at very low abundance. Salvinorins A through D, which are believed to
be produced for protection of the plant, are located in the peltate glandular trichomes that appear
 

1 
 

Figure 1.1: Chemical structure of salvinorin A.

 

2 
 

as hair-like structures mostly on the leaves, but also on external stem tissues, and various other
parts of the plant [8, 11]. Other compounds found in S. divinorum include salvincins A and B,
loliolide, neophytadiene, stigmasterol, oleanolic acid, presqualene alcohol, peplusol, (E)-phytol,
sugars, fatty acids, vitamins, plant sterols, alkanes, and tocopherol (vitamin E) [3, 12, 13].
Geographical location, environmental conditions, and age can affect the composition and
concentration of these compounds [11, 14]. While numerous compounds have been identified,
the entire chemical composition of S. divinorum is unknown [1, 2, 6, 11].

1.2 Use, Abuse, and Effects of Salvia divinorum
S. divinorum has previously been used by the Mazatec Indians for treatment of arthritis,
headaches, and problems with excretion [11]. S. divinorum may be helpful for treatment of
Alzheimer’s, schizophrenia, and stimulant dependence; however, in the United States, there are
currently no accepted medical uses [8, 12]. More recently, S. divinorum has been used
recreationally.
The 2011 National Survey on Drug Use and Health Report approximated 175,000 people
aged 12 or older used S. divinorum in their lifetime [15]. Methods of administration include
chewing, smoking, and ingesting liquid extracts of S. divinorum [2]. Salvinorin A, the active
compound, is absorbed through the oral mucosa, and eventually reaches the brain where it binds
to the kappa opioid receptors causing several effects, which are similar to those caused by
lysergic acid diethylamide (LSD) [8, 6, 11, 12, 16–18]. These effects include hallucinations,
dizziness, slurred speech, and/or uncontrollable laughter and body movements. The onset of the
effects can occur within seconds and, depending on the dose and route of administration, the

 

3 
 

effects may last from minutes to an hour [4, 17]. However, there are currently no known toxic
effects of salvinorin A [17].

1.3 Regulation
As a result of its recent recreational use as a hallucinogenic drug, several countries, which
include Australia, Belgium, Denmark, Estonia, Finland, Italy, Japan, Spain, and Sweden, have
regulated S. divinorum and/or salvinorin A [8, 19]. While the United States Drug Enforcement
Administration (DEA) has included Salvia divinorum and salvinorin A on its Drugs and
Chemical of Concern, they are not currently federally regulated [1, 4, 5, 7]. However, Salvia
divinorum and/or salvinorin A are currently regulated in half of the United States [1, 5, 7, 19,
20]. As of October 1, 2010, S. divinorum and salvinorin A are regulated as schedule I substances
in Michigan, which indicates that these substances have potential for abuse and no accepted
medical use.

1.4 Extraction and Analysis of Salvia divinorum
1.4.1 Extraction Procedures
Due to its recent regulation in several states, the ability to identify S. divinorum and
differentiate it from other plant materials is imperative in a forensic context [5, 9]. However,
various extraction procedures are used to extract salvinorin A from S. divinorum and no
standardized procedure currently exists [1].
Jermain and Evans compared different extraction procedures currently used by forensic
laboratories to extract salvinorin A from S. divinorum [1]. The following extraction procedures
were compared: (1) methanol added to S. divinorum leaves and vortexed at room temperature, as
 

4 
 

used at the Tennessee Bureau of Investigations Central Laboratory in Nashville, (2) 0.1 N
sodium hydroxide and methylene chloride added to leaves, as used at the Missouri State
Highway Patrol Forensic Laboratory, (3) acetone added to S. divinorum leaves for 1 minute, as
used at the Drug Enforcement Administration’s Special Testing and Research Laboratory, (4)
boiling chloroform added to S. divinorum leaves for 10 minutes, as used at the Oklahoma State
Bureau of Investigation, (5) boiling chloroform added to leaves for 1 minute, which was
modified from the above procedure, (6) and chloroform added to leaves at room temperature for
1 minute, which was also modified from the above procedure.
All extracts were analyzed using gas chromatography-mass spectrometry (GC-MS) [1].
Since S. divinorum is typically identified based on the presence of salvinorin A, the extraction
procedures were compared based on the abundance of salvinorin A and the number of additional
plant compounds extracted. The presence of salvinorin A was confirmed by comparing the mass
spectrum from the extract to that of a salvinorin A standard. Of the six procedures, extracting
with chloroform or acetone for one minute was determined to be optimal. In these extracts,
maximum abundance of salvinorin A was extracted with minimal number of additional plant
compounds. However, salvinorin A can be extracted from S. divinorum to prepare liquid extracts
for recreational use [2, 6, 12]. This could result in low levels of salvinorin A in the residual plant
material that may or may not be detectable, causing difficulties with identification. Hence, an
alternative procedure for identifying S. divinorum based on extracting several plant compounds
would be beneficial.

 

5 
 

1.4.2 Analysis Procedures

In the previous study, all extracts were analyzed using GC-MS, which is one of the most
common separation techniques used by forensic laboratories for controlled substance analysis.
However, GC analysis is limited to the analysis of thermally stable, volatile compounds. Liquid
chromatography can be used to analyze non-volatile, thermally unstable compounds that may not
be suitable for GC analysis. For example, liquid chromatography (LC) with ultraviolet (UV)
detection at 208 nm has also been used to analyze extracts of S. divinorum. Gruber et al.
analyzed several S. divinorum extracts using this technique [21]. Based on the retention time of a
reference standard, salvinorin A was detected in all of the extracts. Extracts of various other
Salvia species (i.e. S. concolor, S. blepharophylla, S. chiapensis, S. gregii, S. leucantha, S.
membranacea, and S. recurva) were also analyzed but salvinorin A was not detected in any of
these species. However, LC-UV can only be used for screening purposes since UV detection
does not provide definitive identification of compounds.
High performance liquid chromatography-electrospray ionization-tandem mass
2

spectrometry (LC-ESI-MS ) can provide detailed structural information of chemical compounds;
therefore, compounds can be definitively identified. Tsujikawa et al. analyzed commercially
2

available products of S. divinorum (i.e. leaf and concentrated extract products) by LC-ESI-MS ,
definitively identifying salvinorins A and B in each based on comparison with reference spectra
and subsequent spectral interpretation [6]. Plant extracts of several Salvia species and varieties
(i.e. S. elegans “Scarlet Pineapple”, S. forskaohlei, S. “Indigo Spires”, S. glutinosa, S.
guaranitica, S. blepharophylla, S. Mexicana “Limelight”, S. buchananii, S. splendens, and S.
2

farinacea) were also analyzed using LC-ESI-MS , but neither salvinorin A nor salvinorin B was

 

6 
 

identified. While definitive identification of salvinorin A in Salvia products was demonstrated, a
procedure for identifying S. divinorum based on several plant compounds rather than salvinorin
A alone would be helpful, as explained previously. To do this, additional Salvia species need to
be analyzed and compared to S. divinorum using complete total ion chromatograms (TICs) and
mass spectra.
As previously described, several researchers have demonstrated successful identification
2

of salvinorin A in S. divinorum using GC-MS, LC-UV and LC-MS [1, 6, 21]. Medana et al.
compared the limits of detection of GC-MS, LC-UV, and liquid chromatography-multistage
n

2

3

mass spectrometry (LC-MS ), which included LC-MS and LC-MS

, for the detection of

salvinorins A through D in reference standards [3]. Other compounds, which included
salvinorins E and F, were analyzed using LC-UV to investigate the separation of salvinorins and
n

using LC-MS to determine fragment ions to assist in identification of these compounds.
n

Standards of divinatorins A through C were also analyzed using LC-MS . However, none of the
afore-mentioned compounds were quantified. LC-UV not only had the poorest limit of detection
(i.e. 367 ng/mL for salvinorin A) of the instrumental techniques investigated, but also the UV
n

detector did not allow definitive identification of compounds. Both GC-MS and LC-MS provide
2

structural information so that definitive identification is possible; however, LC-MS and LC3

MS had lower limits of detection for salvinorins A through D than GC-MS (e.g., 3 and 9 ng/mL
2

3

for salvinorin A using LC-MS and LC-MS , respectively compared to 40 ng/mL using GCMS). Therefore, due to improved sensitivity, LC-MS would be more useful than GC-MS in cases
where the salvinorins are present in low concentration. However, other objective comparisons
 

7 
 

may be beneficial, such as the use of statistical procedures to evaluate and compare these
techniques.

1.4.3 Statistical Procedures
Both GC and LC have been previously used to analyze extracts of Salvia species.
However, the data produced from these analyses are very complex and comparison of
chromatograms can be difficult and even subjective. In 2009, the National Academy of Sciences
published a report titled “Strengthening Forensic Science in the United States: A Path Forward,”
which stated 13 recommendations to improve the field of forensic science [22]. Of these
recommendations, one stated the need for more research to evaluate procedures, especially those
that may be subjective. Therefore, more objective procedures, such as the use of statistical
procedures, may be beneficial in forensic analyses [19, 22].
Statistical analyses have been previously applied to Salvia data. Zhou et al. extracted and
analyzed 25 samples of 12 Salvia species (S. divinorum was not included in this data set) by
ultra-high-performance liquid chromatography-tandem mass spectrometry [23]. The resulting
chromatograms were subjected to various pretreatment procedures (i.e. peak finding, peak
alignment and peak filtering) to minimize non-chemical sources of variance in the data, prior to
statistical analysis. Principal component analysis (PCA) was then performed to determine which
compounds could be used as chemical markers for the different species. In the resulting scores
plot, the 12 species formed three clusters. The first cluster included samples extracted from the
roots of plants, while the second cluster included samples extracted from the aerial parts of the
plants. The third cluster included two samples: one extracted from the aerial part of the plant and
one extracted from the whole plant. There was a number of different species within each cluster.
 

8 
 

Although three clusters were apparent to the authors, it does not appear that any statistical
assessment of the scores plot was performed to determine this clustering. According to the
loadings plot, dihydrotanshinone I was the most variable compound and, thus, could be used as a
chemical marker to differentiate samples extracted from roots of plants and those extracted from
the aerial parts of the plants.
Carrer et al. also used PCA to assess Salvia data [14]. Leaves of S. guaranitica from
seven different locations in Brazil were analyzed by GC-MS. Several compounds in S.
guaranitica were identified, which included α-pinene, β-cubebene, β-caryophyllene, δ-cadinene,
caryophyllene oxide, and phytol. The presence of the compounds varied based on the
geographical origin of S. guaranitica. In the PCA scores plot, the S. guaranitica samples were
grouped in three clusters based on variation in composition. However, no additional statistical
assessment of the positioning of the clusters was performed.
Bodnar Willard et al. were the first to use PCA, along with various other statistical
procedures, to assess the discrimination of S. divinorum from related Salvia species (i.e. S.
guaranitica, S. nemorosa, S. officinalis, and S. splendens) based on TICs obtained using GC-MS
[19]. To minimize non-chemical sources of variance, the data were background corrected,
smoothed, retention time aligned, and normalized prior to statistical analysis. Using PCA, S.
divinorum was easily discriminated from the other Salvia species through visual assessment of
the scores plot.
Several additional statistical procedures were then conducted to provide a more objective
interpretation of positioning of the Salvia species on the scores plot [19]. Euclidean distances
were calculated between S. divinorum and each of the other species, as well as between
replicates of the same species, based on the mean scores. Using this metric, S. divinorum was
 

9 
 

distinct from the other species since the Euclidean distance was 4.7–17.5 times greater than the
distance observed between replicates of the same species. Student’s t-tests and Wilcoxon ranksum tests were also performed on the mean PCA scores for each species. Using both tests, S.
divinorum was statistically differentiated from the other Salvia species on both principal
component 1 (PC1) and principal component 2 (PC2), at the 99.9 % confidence level.
Hierarchical cluster analysis (HCA) was also performed on the mean PCA scores to assess
similarity, or otherwise, of the species. This procedure showed that there was greater similarity
among replicates of the same species (similarity index = 0.776–0.983) than between S. divinorum
and the other Salvia species (similarity index = 0.000–0.528). Finally, to further assess
similarity, PPMC coefficients were calculated between replicates of the same species, as well as
between S. divinorum and the other Salvia species. Specifically, the coefficients were calculated
for pair-wise comparisons of the loadings of each sample on PC1. Mean PPMC coefficients (r =
0.9477–0.9981) indicated that replicates were strongly correlated, while S. divinorum was
weakly correlated to the other Salvia species (r = 0.2906–0.4428). Therefore, S. divinorum was
differentiated from the other Salvia species using all these procedures, which offered an
objective interpretation of the scores plot.
Although Bodnar Willard et al. demonstrated statistical discrimination of S. divinorum
from other Salvia species based on the full TICs, the extraction procedure was optimized for the
extraction of salvinorin A, rather than the entire chemical fingerprint [19]. In addition, the
extracts were only analyzed by GC-MS, which is limited to the analysis of thermally stable,
volatile compounds. As previously mentioned, LC-MS is generally more sensitive and can be
used to analyze non-volatile compounds. Therefore, LC-MS may provide additional chemical
information about the plant material.
 

10 
 

1.5 Identification of other Salvia species
Over 900 Salvia species exist and, hence, the ability to identify S. divinorum and
differentiate it from other plant material is imperative in a forensic context [5, 9]. Several Salvia
species, which include S. guaranitica, S. nemorosa, S. officinalis, and S. splendens, have been
analyzed previously using various analytical procedures such as gas chromatography-flame
2

ionization detection (GC-FID), GC-MS, LC-MS , and

13

CNMR [24–27]. Many compounds

from these species have been identified, some of which are common among species, including S.
divinorum.

1.5.1 S. guaranitica
S. guaranitica, a plant used for decorative and medicinal purposes, belongs to the same
clade as S. divinorum, clade II [7, 9, 14, 28]. Vallverdú et al. determined the chemical
composition of S. guaranitica using GC-FID, GC-MS, and

13

CNMR to investigate possible

medicinal uses [24]. A total of 49 compounds were identified, which included mostly
sesquiterpenes, oxygenated sesquiterpenes, and monoterpenes. Phytol, an oxygenated diterpene,
previously found in S. divinorum was also identified in S. guaranitica [3, 13, 14, 24].

1.5.2

S. nemorosa
S. nemorosa, a plant used for decorative purposes, belongs to clade I [7, 9, 28]. Malenčić

et al. analyzed S. nemorosa by GC-FID and GC-MS to determine its chemical composition [25].
A total of 23 compounds were identified, which included monoterpenes, sesquiterpenes, and
other hydrocarbons. Some of these compounds included α-pinene, α-cadinol, α-copaene, α 

11 
 

humulene, p-cymene, δ-cadinene, and caryophyllene oxide, which have also been identified in S.
guaranitica. Takeda et al. also identified salvinosides A through C in S. nemorosa by UV,
1

HNMR,

13

CNMR, negative ion fast atom bombardment-mass spectrometry, and circular

dichroism spectroscopy [26]. Ulubelen et al. identified nemorosin and salvinemorol in S.
1

nemorosa by UV spectroscopy, IR spectroscopy, HNMR spectroscopy, and thin layer
chromatography (TLC) [29]. Other compounds present in S. nemorosa included oleanolic acid,
which was previously found in S. divinorum, sitosterol, and ursolic acid [3, 12, 13, 29].

1.5.3

S. officinalis
S. officinalis, an herb that has been researched greatly due to its medicinal properties,

belongs to clade I [7, 9, 24, 30–32]. This plant material has antiviral, antitumor, and antibacterial
properties [27, 30, 33]. Radulescu et al. analyzed S. officinalis by GC-MS to develop an
extraction procedure and determine its chemical composition [30]. A total of 35 compounds were
identified, some of which were common to other Salvia species. For example, β-pinene, βcaryophyllene, and β-elemene were also identified in S. guaranitica, while p-cymene and αhumulene were also found in S. guaranitica and S. nemorosa. In addition, oleanolic acid was
identified in S. nemorosa and S. divinorum. Vitamin E was also found in S. divinorum. In a
review of the polyphenolic compounds in Salvia, Lu et al. reported that cirsiliol was identified in
both S. officinalis and S. guaranitica [32]. Djarmati et al. extracted S. officinalis and identified
manool, rosmanol 7-ethyl ether, oleanolic acid, and carnosic acid 12-methyl ether-γ-lactone
1

using combinations of IR, HNMR,

13

CNMR, TLC, and X-ray diffraction [34]. Abreu et al.

extracted and analyzed leaves of 60 different Salvia species, including S. officinalis, using LC 

12 
 

2

MS and identified carnosic acid, carnosol, methoxycarnosic acid, isorosmanol, and vitamin E
[33]. Kivilompolo et al. analyzed S. officinalis by GC-MS, LC-UV, and LC-MS to compare
qualitative and quantitative results of specific compounds [27]. Several compounds were
detected; however, all compounds were not detected by every technique. For instance,
rosmarinic acid was detected by LC-UV and LC-MS, but not by GC-MS. Other compounds that
have also been identified in S. officinalis include α-thujone, β-thujone, camphor, and 1-borneol
[31].

1.5.4 S. splendens
S. splendens, a plant used for decorative purposes, belongs to the same clade as S.
divinorum, clade II [7, 9, 28]. Several compounds have been identified in S. splendens, some of
which are structural congeners of salvinorin A [35]. These compounds included salvisplendins A
through D, salviarin, splendidin, and splenolides A and B, which were extracted from S.
1

splendens and analyzed using a combination of techniques such as IR, UV, HNMR,

13

CNMR,

and MS.

1.6 Objective
With the limitations in previous studies, the objective of this research was to investigate
the association and discrimination of S. divinorum from other Salvia species (i.e. S. guaranitica,
S. nemorosa, S. officinalis, and S. splendens) based on both volatile and non-volatile compounds.
In order to achieve this objective, several goals had to be met. First, all species were extracted
and analyzed using GC-MS and LC-MS to generate chemical fingerprints based on the volatile
and non-volatile compounds, respectively. Compounds present in the TICs were tentatively
 

13 
 

identified using available MS library databases and literature. Data pretreatment procedures were
performed to minimize non-chemical sources of variance. To assess the association and
discrimination of S. divinorum from the other four Salvia species, PCA was performed separately
on the GC and LC data. To objectively evaluate the PCA scores plot, Student’s t-tests were
performed on the mean scores of each species. In addition, PPMC coefficients were calculated
between TICs to further assess this association and discrimination. The results from the statistical
analyses of the GC and LC data were then compared to determine the analytical technique that
provided more discriminating information for S. divinorum. Since there are many Salvia species,
the ability to discriminate S. divinorum from related Salvia species is important in a forensic
laboratory. Currently, S. divinorum is identified by the presence of salvinorin A; however,
salvinorin A can be extracted from the leaf material. Therefore, an alternative procedure to
identify S. divinorum based on several compounds may be useful.

 

14 
 

References

 

15 
 

References
[1] Jermain JD, Evans HK. Analyzing Salvia divinorum and its active ingredient salvinorin A
utilizing thin layer chromatography and gas chromatography/mass spectrometry. J Forensic Sci
2009; 54:1–5.
[2] Medana C, Massolino C, Pazzi M, Baiocchi C. Determination of salvinorins and divinatorins
in Salvia divinorum leaves by liquid chromatography/multistage mass spectrometry. Rapid
Commun Mass Sp 2006; 20:131–6.
[3] Giroud C, Felber F, Augsburger M, Horisberger B, Rivier L, Mangin P. Salvia divinorum: an
hallucinogenic mint which might become a new recreational drug in Switzerland. Forensic Sci
Int 2000; 112:143–50.
[4] Lange JE, Reed MB, Ketchie Croff JM, Clapp JD. College student use of Salvia divinorum.
Drug and Alcohol Depen 2008; 94:263–6.
[5] McDonough PC, Holler, JM, Vorce SP, Bosy TZ, Magluilo J Jr, Past MR. The detection and
quantitative analysis of the psychoactive component of Salvia divinorum, Salvinorin A, in human
biological fluids using liquid chromatography-mass spectrometry. J Anal Toxicol 2008; 32:417–
21.
[6] Tsujikawa K, Kuwayama K, Miyaguchi H, Kanamori T, Iwata YT, Yoshida T, Inoue H.
Determination of salvinorin A and salvinorin B in Salvia divinorum-related products circulated
in Japan. Forensic Sci Int 2008; 180:105–9.
[7] Bodnar Willard MA. Forensic analysis of Salvia divinorum and related Salvia species using
chemometric procedures. MS Thesis (2010).
[8] Drug Enforcement Agency. Salvia divinorum and salvinorin A. 2010. www.deadiversion.
usdoj.gov/drugs_concern/salvia_d.pdf
[9] Walker JB, Sytsma KJ, Treutlein J, Wink M. Salvia (Lamiaceae) is not monophyletic:
implications for the systematics, radiation, and ecological specializations of Salvia and tribe
mentheae. Am J Bot 2004; 91:1115–25.
[10] Ortega A, Blount JF, Manchand PS. Salvinorin, a new trans-neoclerodane diterpene from
Salvia divinorum (Labiatae). J Chem Soc Perkin Trans 1 1982; 2505–8.
[11] Siebert, DJ. Localization of salvinorin A and related compounds in glandular trichomes of
the psychoactive sage, Salvia divinorum. Annals of Botany 2004; 93:763–71.
[12] Grundmann O, Phipps SM, Zadezensky I, Butterweck V. Salvia divinorum and salvinorin
A: An update on pharmacology and analytical methodology. Planta Med 2007; 73:1039–46.

 

16 
 

[13] Bigham AK, Munro TA, Rizzacasa MA, Robins-Browne RM. Divinatorins A–C, new
neoclerodane diterpenoids from the controlled sage Salvia divinorum. Journal of Natural
Products 2003; 66:1242–4.
[14] Carrer RP, Vanderlinde R, Dutra S, Marcon A, Echeverrigaray S. Esstential oil varation
among Brazilian accessions of Salvia guaranitica L.. Flavour Fragr. J. 2007; 22:430–4.
[15] Results from the 2011 national survey on drug use and health: detailed tables. 2011.
http://www.samhsa.gov/data/NSDUH/2011SummNatFindDetTables/NSDUHDetTabsPDFWHTML2011/2k11DetailedTabs/Web/HTML/NSDUH-DetTabsSect7seTabs1to452011.htm#Tab7.28C
[16] Kennedy JH, Wiseman JM. Direct analysis of Salvia divinorum leaves for salvinorin A by
thin layer chromatography and desorption electrospray ionization multi-stage tandem mass
spectrometry. Rapid Commun Mass Sp 2010; 24:1305–11.
[17] Lange JE, Daniel J, Homer K, Reed MB, Clapp JD. Salvia divinorum: Effects and use
among YouTube users. Drug and Alcohol Depen 2010; 108:138–40.
[18] Drug Enforcement Administration, Office of Diversion Control, Drug and Chemical
Evaluation Section. Salvia divinorum and Salvinorin A. December 2010.
[19] Bodnar Willard MA, McGuffin VL, Waddell Smith R. Forensic analysis of Salvia
divinorum using multivariate statistical procedures. Part I: discrimination from related Salvia
species. Anal Bioanal Chem 2012; 402:833–42.
[20] Enrolled House Bill No. 6038. State of Michigan 95th Legisature, Regular Session. 111
(2010).
[21] Gruber JW, Siebert DJ, Der Marderosian AH, Hock RS. High performance liquid
chromatographic quantification of salvinorin A from tissues of S. divinorum Epling & Játiva-M.
Phytochem. Anal. 1999; 10:22–5.
[22] Committee on identifying the needs of the forensic sciences community, National Research
Council. Strengthening Forensic Science in the United States: A Path Forward. 2009.
https://www.ncjrs.gov/pdffiles1/nij/grants/228091.pdf
[23] Zhou Y, Xu G, Fung Kei Choi F, Ding L, Bin Han Q, Song J, Feng Qiao C, Zhao Q, Xu H.
Qualitative and quantitative analysis of diterpenoids in Salvia species by liquid chromatography
coupled with electrospray ionization quadrupole time-of-flight tandem mass spectrometry.
Journal of Chromatography A 2009; 1216:4847–58.
[24] Vallverdú C, Vila R, Lorenzo D, Paz D, Dellacassa E, Davies P, Villamil J, Tomi F,
Casanova J, Cañigueral S. Composition of the essential oil of cultivated Salvia guaranitica from
Uruguay. Flavour Fragr. J. 2005; 20:421–4.
 

17 
 

[25] Malenčić D, Couladis M, Mimica-Dukić N, Popović M, Boža P. Essential oils of three
Salvia species from the Pannonian part of Serbia. Flavour Fragr. J. 2004; 19:225–8.
[26] Takeda Y, Zhang H, Matsumoto T, Otsuka H, Oosio Y, Honda G, Tabata M, Fujita T, Sun
H, Sezik E, Yesilada E. Megastigmane glycosides from Salvia Nemorosa. Phytochemistry 1997;
44:117–20.
[27] Kivilompolo M, Obůrka V, Hyötyläinen T. Comparison of GC-MS and LC-MS methods for
the analysis of antioxidant phenolic acids in herbs. Anal Bioanal Chem 2007; 388:881–7.
[28] Tychonievich, J. Salvia Clades. Personal Communication. 2009.
[29] Ulubelen A, Topcu G, Sönmez U, Eris C. Terpenoids from Salvia nemorosa.
Phytochemistry 1994; 35:1065–7.
[30] Kan Y, Gökbulut A, Kartal M, Konuklugil B, Yilmaz G. Development and validation of a
LC method for the anlaysis of phenolic acids in Turkish Salvia species. Chromatographia 2007;
66:S147–52.
[31] Radulescu V, Chiliment S, Oprea E. Capillary gas chromatography-mass spectrometry of
volatile and semi-volatile compounds of Salvia officinalis. Journal of Chromatography A 2004;
1027:121–6.
[32] Lu Y, Yeap Foo L. Polyphenolics of Salvia–a review. Phytochemistry 2002; 59:117–40.
[33] Abreu ME, Müller M, Alegre L, Munné-Bosch S. Phenolic diterpene and α-tocopherol
contents in leaf extracts of 60 Salvia species. J Sci Food Agric 2008; 88:2648–53.
[34] Djarmati Z, Jankov RM, Djordjević A, Ribar B, Lazar D, Engel P. Carnosic acid 12-methyl
ether-γ-lactone, a ferruginol-type diterpene from Salvia officinalis. Phytochemistry 1992;
31:1307–9.
[35] Fontana G, Savona G, Rodríguez B, Dersch C, Rothman RB, Prisinzano TE. Synthetic
studies of neoclerodane diterpenoids from Salvia splendens and evaluation of opioid receptor
affinity. Tetrahedron 2008; 64:10041–8.

 

18 
 

Chapter 2: Theory
2.1 Instrumental Techniques
Chromatography is a separation technique, which is widely used in forensic science for a
variety of applications including, but not limited to, the analysis of controlled substances [1]. The
goal of any chromatographic technique is to separate the sample mixture into individual
compounds [2]. To do this, all chromatographic techniques contain a stationary and a mobile
phase. The stationary phase is either a solid or a liquid film, which can have different chemical
compositions. There are many polar and non-polar stationary phases that can be used and the
selection of the stationary phase is dependent on the nature of compounds to be separated [3, 4].
While the stationary phase is static, the mobile phase, which is composed of either a gas or a
liquid, is continuously flowing and carries the sample through the instrument [5]. Compounds in
a sample mixture will interact with the stationary and mobile phases to achieve separation. In this
research, both gas and liquid chromatography coupled with mass spectrometry was used.

2.1.1 Gas Chromatography-Mass Spectrometry
In modern gas chromatography (GC), which is used to analyze thermally stable and
volatile compounds, the stationary phase is a liquid film, while the mobile phase is a pressureregulated gas, more commonly referred to as the carrier gas [3–5]. In addition to interactions
between phases, separation in GC is also based on the boiling points of the compounds in the
sample mixture. While solid, liquid, and gaseous samples can be analyzed using GC, typically in
controlled substance analysis, the solid sample is dissolved in a suitable solvent, which generates
a liquid for analysis. The sample is injected via syringe through a rubber septum into the GC

19
 

(Figure 2.1), which consists of a heated inlet, an oven containing a column coated with the
stationary phase, and a detector, which was a mass spectrometer in this research [1, 3].
Upon injection, the sample is volatilized in the heated inlet, which contains a split valve
that can be opened or closed during the injection [5]. If the valve is closed during injection, this
is called a splitless injection [4]. Under these circumstances, all of the injected sample enters the
column. As a result, splitless injections are used to analyze samples at low concentrations [1].
However, when using a splitless injection, compounds may thermally degrade due to the time
they spend in the inlet. Alternatively, if the valve is open during injection, this is called a split
injection, in which case, only a portion of the sample enters the column. For example, a split
ratio of 50:1 indicates that for every 1 part of sample that enters the column, 50 parts are vented
to waste. In particular, the higher molecular weight compounds are lost since they do not
volatilize as easily as the lower molecular weight compounds and, hence, this type of injection
may not be fully representative of the sample [4]. Split injections are used to analyze highly
concentrated or dirty samples.
The carrier gas flows throughout the system at a rate ranging from 0.5 to 2 mL/min. The
role of the carrier gas is to carry the volatilized sample onto the column. Common carrier gases
include nitrogen, helium, and hydrogen [2, 4, 5]. However, high purity helium is primarily used
for gas chromatography-mass spectrometry (GC-MS) analysis because it is light and inert [2, 5].
Therefore, it is efficiently pumped away by the mass spectrometer vacuum system, which
enables the direct connection of the gas chromatograph and mass spectrometer.
Ideally, the sample mixture is injected onto the column in as narrow of a band as possible
[3, 4]. The column is capillary tubing with a thin coating of the liquid stationary phase on the

20
 

Inlet

Transfer
Line
Capillary
Column

Carrier
Gas
Oven

Figure 2.1: Schematic diagram of a gas chromatograph.

21
 

Detector

inner walls allowing the carrier gas to flow through the center. As previously stated, there are
many stationary phases with different chemical compositions that are used for different
applications. For example, a non-polar stationary phase composed of a polysiloxane backbone
with 5% phenyl and 95% methyl substitutes is useful when separating non-polar compounds
such as hydrocarbons [4]. Sample compounds partition between the carrier gas and the stationary
phase and, since the carrier gas is inert, the interactions between the compounds in the sample
and the stationary phase lead to separation [1, 4, 5]. Possible interactions that may occur include
dispersion forces, dipole induction forces, dipole orientation forces, and hydrogen bonding [4].
Therefore, separation is affected by properties of the compounds, such as size, polarizability, and
acidity or basicity.
The efficiency of the separation is affected by the specifications of the column, mainly
length (typical range 10–30 m), internal diameter (typical range 0.1–0.32 mm), and stationary
phase thickness (typical range 0.25 or 0.32 µm) [2, 4, 5]. In an efficient separation, compounds
should travel through the column in as narrow a band as possible, resulting in narrow, Gaussianshaped peaks in the final chromatogram [4].
Narrower internal diameters can also result in a more efficient separation [2]. A capillary
column with a narrow internal diameter allows the carrier gas to flow through the column
unrestricted without major pressure drops. As a result, longer capillary columns are used, which
provide more time for interactions between the compounds and the stationary phase and thus
more separation. However, longer time spent in the column leads to greater broadening from
molecular diffusion. Molecular diffusion occurs when molecules of the same compound diffuse
from regions of high concentration to regions of low concentration in the mobile and stationary
phases [4, 5]. Due to molecular diffusion, band broadening increases with time.

22
 

A thinner stationary phase can also provide a more efficient separation due to rapid mass
transfer [5]. Mass transfer occurs when molecules of a compound move between the mobile and
stationary phases [4, 5]. To avoid band broadening, ideally the molecules in the mobile and
stationary phases should be in equilibrium with one another. However, since the mobile phase is
continuously moving, the system is dynamic, and thus equilibrium must be re-established
constantly. The faster this is achieved, the less band broadening occurs; however, sufficient time
for separation is still needed. Therefore, there must be a compromise between column length,
internal diameter, and thickness of the stationary phase to obtain an efficient separation,
To further control the separation and minimize band broadening, the temperature of the
oven can be adjusted. The oven, in which the column is located, can be set to maintain a constant
temperature for an entire analysis, which is called an isothermal analysis, or to vary temperature
throughout the analysis, which is called temperature programming [3–5]. This type of
programming allows for the separation of complex mixtures with a wide range of boiling points.
Therefore, temperature programming can decrease analysis time, which limits molecular
diffusion, decreases mass transfer time, and minimizes band broadening. However, a decrease in
analysis time can result in a loss in resolution and, thus, there must be a compromise between
analysis time and resolution.
Many different detectors are available for GC, each of which generates a chromatogram
of the sample [2]. This is a plot with retention time, which is the time it takes for a compound to
travel through the column and reach the detector, on the x-axis and abundance on the y-axis [1,
3]. Each detected compound is represented by a peak in the chromatogram with a specific
abundance at a given retention time.

23
 

Despite the number of detectors available, specifically in this research, a mass
spectrometer (MS) detector was used [2]. The GC column passes through a transfer line into the
mass spectrometer. The transfer line is typically heated to the equivalent of the highest GC oven
temperature to prevent the loss of separated compounds due to condensation. The mass
spectrometer is maintained under vacuum (approximately 10-4 to 10-7 Pa) to reduce ion collisions
with gas molecules, to increase the mean free path of the ions, and to protect the mass
spectrometer from moisture that would cause corrosion and degrade performance [1, 3]. Due to
the low flow rates used with capillary GC columns and the low molecular weight of helium, the
carrier gas is efficiently pumped away by the MS vacuum system. Therefore, it is possible to
directly couple the GC to the MS through the transfer line, avoiding the need for a specialized
interface [2].
The mass spectrometer consists of an ionization source, a mass analyzer, and a detector
[3]. The source ionizes the separated compounds from the GC column and fragments those ions.
The mass analyzer separates the ions according to mass-to-charge (m/z) ratio, and the detector
converts the ions into an electric signal to be detected.
Many types of ionization methods are available but, in this research, electron ionization
(EI) was used. Figure 2.2 shows a schematic diagram of an EI source. During EI, a filament
produces an electron beam (typically 70 eV) that is attracted to a positive plate on the opposite
side of the source [1, 3]. Due to the presence of a magnet in the source, the electron beam travels
in a spiral path, which increases the chance of interaction between the beam and the separated
compounds from the GC column. The separated compounds pass by the beam perpendicularly
and the electron beam causes the ejection of an electron from the compounds, leaving them with
a positive charge [1, 3]. This ion has the same molecular mass as the original compound and is

24
 

Magnet
Filament

Positive
Repellor
Plates
+

Electron
Beam
+ + + + + + +
+

Separated Compounds
Transfer
Line

+

Positive
Plate
Magnet

Figure 2.2: Schematic diagram of an electron ionization source adapted from [1].

25
 

To
Mass
Analyzer

known as the molecular ion [1]. The amount of energy needed to ionize a compound, known as
the ionization energy, varies based on the structure of the compound; however, approximately 5
to 25 eV is a sufficient amount to ionize most compounds [6]. Therefore, the excess energy
causes the molecular ion to fragment. The way in which the ion fragments is unique to that ion
and will occur in a reproducible manner under the same ionization conditions. Thus, EI is
considered a hard ionization method due to the extent of fragmentation that occurs.
The ions formed are then repelled by positive repellor plates into the mass analyzer,
which separates the ions based on their m/z value [3]. Several different types of mass analyzers
are available but, in this research, a quadrupole analyzer was used (Figure 2.3). The quadrupole
consists of four rods with opposite rods connected as a pair. Alternating direct current (DC) and
radio frequency (RF) voltages are applied to the two pairs of rods. A specific ratio of DC and RF
voltages will allow only ions with a narrow range of m/z values to pass through the quadrupole
unhindered and reach the detector [1, 3]. Ions with a m/z value outside this range will hit the rods
of the quadrupole, be neutralized, and pumped away by the vacuum system [3]. The quadrupole
can operate in full scan mode, during which both the DC and RF voltages are increased while
still maintaining the same DC/RF ratio [1]. At each DC/RF voltage, ions of different m/z values
will pass through the quadrupole to the detector. In this way, a wide range of m/z values will
reach the detector (e. g. m/z 50–550) to obtain a full scan.
In bench-top GC-MS instruments, continuous dynode electron multipliers are commonly
used as the detector (Figure 2.4). This detector is composed of a curved glass tube that is coated
with a substance (e.g. lead oxide) that readily emits secondary electrons [1, 2, 5]. The positively
charged ions from the mass analyzer hit the surface of the electron multiplier, causing ejection of
electrons from the multiplier surface [1, 3]. A potential gradient exists along the length of the

26
 

+
-

Ion
+

Alternating
Direct Current
or
Radio Frequency
Voltages

Figure 2.3: Schematic diagram of a quadrupole mass analyzer adapted from [1, 2].

27
 

+

Coated With Conductive Substance
Ions

+

+

Ions
+

+

+

Electrons

Figure 2.4: Schematic diagram of a continuous dynode electron multiplier adapted from [1].

28
 

multiplier, attracting electrons down the multiplier toward the ground potential [1]. As the
electrons travel, further collisions with the surface occur, which causes ejection of more
n

electrons [1, 3]. For every n collisions with the surface, the signal is amplified by 2 . When the
cascade of electrons reaches the end of the multiplier, the signal has been amplified by
approximately 105 to 108 [1, 2, 5].
The current is measured and digitized, producing the mass spectrum of the compound.
The spectrum is a plot of ion abundance versus m/z value. In many cases, the ion with the highest
m/z value is the molecular ion, which is useful in identifying the compound’s molecular mass.
Every mass spectrum also has a base peak, which is the most abundant ion. In addition, the
masses of the fragment ions are used to determine the structure of the compound. Figure 2.5
shows the mass spectrum of salvinorin A, which is the hallucinogenic compound in Salvia
divinorum, generated using EI with the molecular ion, base peak, and major fragments labeled
[7, 8].
As previously stated, GC can be coupled with several detectors; however, these detectors
are not specific for identification purposes since only a chromatogram, and hence retention time
information, is obtained for each compound in the sample mixture [2]. Therefore, only retention
times between compounds can be compared. In GC-MS, the chromatogram is more correctly
called the total ion chromatogram (TIC), since the chromatogram is plotted as the sum of all the
ion signals in the sample [3]. Using GC-MS, every compound in the TIC has a given retention
time and a mass spectrum; thus, definitive identification of unknown compounds is possible.
Therefore, GC-MS is widely used in forensic science to identify unknown substances.

29
 

7.00E+05

Intensity

Base Peak
94

273
121

166 220

318 359

Molecular Ion
432
404

0.00E+00
50 100 150 200 250 300 350 400 450 500
Mass to Charge (m/z)
Figure 2.5: Mass spectrum of salvinorin A analyzed using electron ionization.

30
 

2.1.2 Liquid Chromatography-Mass Spectrometry and Tandem Mass Spectrometry
Currently, liquid chromatography-mass spectrometry (LC-MS) is not routinely used in
forensic science; however, it has application for separation and identification of non-volatile and
thermally degradable compounds that may not be detected by GC-MS [4]. A liquid
chromatography system (Figure 2.6) consists of a fixed-volume injection loop, a liquid mobile
phase, pumps for the mobile phase, a column containing a packed solid stationary phase, and a
detector.
Since the sample is injected onto the column as a liquid, samples not in liquid form must
first be prepared. Once prepared, the sample is drawn into a syringe and then injected into the
fixed-volume injection loop [4, 5]. The injection loop typically contains 1 to 25 µL of sample in
the ‘load’ position; however, it can contain up to 100 µL of sample [2]. The loop is then placed
into the ‘inject’ position, during which the mobile phase enters the loop and carries the sample
onto the column [2, 4, 5]. Fixed volume injection loops are advantageous since injection volume
is more reproducible, which leads to improved analytical precision [2].
Columns in liquid chromatography are typically 10 to 30 cm in length, 4 to 5 mm in
internal diameter, and contain a solid stationary phase with particle sizes ranging from 3 to 5 µm
[4, 5]. Since packed columns are used in LC, peak broadening can result not only from molecular
diffusion and mass transfer, but also from “multiple path” contribution. This occurs when
molecules of the same compound take different paths through the packed particles of the
stationary phase [5]. In order to minimize peak broadening due to “multiple path” contribution, a
smaller particle size can be used, enabling tighter packing and reducing the number of pathways
available. In some LC systems, the column may be used inside an oven [5]. For LC, temperature

31
 

Syringe

Fixed-Volume
Injection Loop
Packed
Column

Pumps
Mobile Mobile
Phase 1 Phase 2

Figure 2.6: Schematic diagram of a liquid chromatograph adapted from [5].

32
 

Detector

programming is not commonly used; instead, a constant oven temperature is used to prevent
temperature variations that could affect retention times [4, 5].
Efficient separation in LC results from interactions between the compounds in the
sample, the stationary phase, and the mobile phase [4, 5]. Several types of interactions may occur
and various properties of the compounds, stationary phase, and the mobile phase affect
separation [4]. However, the separation in LC is ultimately based on polarity and solubility [4,
5]. For example, in normal-phase chromatography, a polar stationary phase and non-polar mobile
phase are used to separate compounds, while in reversed phase chromatography, a non-polar
stationary phase and a polar mobile phase are used [4, 9]. Reversed phase chromatography is
more commonly used for controlled substance analysis and has been shown to separate a variety
of compounds such as depressants and pain killers, which makes it useful for forensic science
analyses [4, 5].
To further assist with separation, the composition of the mobile phase, which can consist
of different solvents with varying polarities, can be programmed to change over time [5]. This is
known as solvent programming, which is analogous to temperature programming used in gas
chromatography. In solvent programming, the concentration of one solvent is programmed to
increase during the analysis while another solvent’s concentration decreases. This consequently
changes the polarity of the mobile phase. It is common in reversed phase chromatography to
decrease the polarity of the mobile phase by decreasing the concentration of the more polar
solvent and increasing the concentration of the less polar solvent [4, 9]. In general, this decrease
in polarity of the mobile phase will gradually cause the non-polar compounds to elute from the
stationary phase after the polar compounds have been separated. Therefore, solvent

33
 

programming allows for the efficient separation of compounds with a wide range of polarities
and also decreases analysis time [5].
The separated compounds then enter the detector. As for GC, a number of different types
of detectors are available for LC; however, in this research, a mass spectrometer was used as the
detector [2]. The column effluent first enters a heated capillary interface, which leads into the
ionization source [3]. While many ionization methods are available, electrospray ionization (ESI)
was used in this research. This is a common ionization method for analysis by LC-MS because
ionization occurs in solution at atmospheric pressure, which makes ionization of non-volatile
compounds possible [1]. Contrary to EI, ESI is a soft ionization method, during which little
fragmentation of the pseudomolecular ion occurs [1, 10]. As a result, the pseudomolecular ion is
typically apparent; however, it is difficult to identify compounds since less fragment ions are
available for structural elucidation [10].
In ESI, ion formation is either the result of compound interaction with the solvent
solution in which it is dissolved or with the mobile phase [1]. ESI can be performed in positive
or negative mode, which results in both singly charge ions, such as the pseudomolecular ion, and
multiply charge ions. In negative mode, ionization occurs by deprotonation or by forming a
negatively charged adduct; therefore, this mode is useful for ionizing acidic compounds [1, 10].
In positive mode, ionization occurs by protonation or by forming a positively charged adduct;
therefore, this mode is useful for ionizing basic compounds. In this research, positive ionization
was used and will be discussed further.
To couple the LC system to the mass spectrometer, the column effluent first enters a
heated capillary interface, which leads into the ionization source [3]. The ESI source (Figure 2.7)
consists of a needle and a counter electrode [1, 2, 10]. A potential difference, known as the ion

34
 

Nebulizing
Gas

Drying
Gas

Needle

Taylor
Cone
v
+ + ‐ + +
+ +
v v +v
+ ‐ + + ‐ ++
+
+
v v ‐
v v
+ + v ++ +
+ +

+
++

+
+ M+H+
++ + +
+

++
+

Separated
Compounds
Flow
of
Electrons

++

Electrode

+ Power
Supply

To
Mass
Analyzer

‐

Figure 2.7: Schematic diagram of electrospray ionization source operating in positive ionization
mode adapted from [1].

35
 

spray voltage, is applied between the two, which generates an electric field [1, 10]. Charge is
accumulated as liquid, which consists of the separated compounds and mobile phase, emerges
from the needle. Due to the potential difference between the needle and counter electrode, the
liquid forms a“Taylor cone” [1, 6]. The source also contains a nebulizing gas, which assists with
droplet formation, and a drying gas, which assists with solvent evaporation [1, 10]. As the
droplets get smaller, the charge density increases and the ions begin to repel one another [1, 3,
10]. Ions in the gas phase are formed as the charged droplets from the Taylor cone undergo
solvent evaporation and Coulomb fissions. The exact mechanism of ionization is unknown;
however, it is commonly thought that these Coulomb fissions occur either by the charge residue
mechanism or the ion evaporation mechanism [10, 11].
In the charge residue mechanism, repulsion forces overcome surface tension forces [6,
10, 11]. As a result, the droplets become smaller and smaller. This process continues until the
droplet contains only one ion. Throughout the mechanism, the drying gas evaporates the solvent
and eventually the ion enters the gaseous state. In the ion evaporation mechanism, the electric
field on the surface of the droplet overcomes surface tension forces. As a result, the ion escapes
the droplet and immediately enters the gaseous state. Then, the ions pass through a capillary into
differentially pumped vacuum regions, and are focused into the mass analyzer [1, 2].
In this research, a hybrid triple quadrupole/linear ion trap (Figure 2.8) was used as the
mass analyzer. This consists of three quadrupoles (Q 0, Q1, and Q2) and a hybrid quadrupole
(Q3), which is a linear ion trap (LIT) [10, 12]. The first quadruple (Q 0) is not used as a mass
analyzer, but is used to focus and collect the ions, which ultimately increases the ion signal [1, 9,
10, 12]. The second quadrupole (Q1) is capable of mass analysis, while the third quadrupole (Q2)

36
 

Q0

Q2

Q1

Linear Ion Trap
Q3

+ +
Ions
+
+ +
Figure 2.8: Schematic diagram of a hybrid triple quadrupole/linear ion trap adapted from [1].

37
 

is used as a collision cell. The LIT is also capable of mass analysis. It consists of four rods,
arranged similarly to a quadrupole, with electrodes at each end [1]. Potentials are applied to these
electrodes, which allow the LIT to confine ions. Linear ion traps have a greater ion trapping
capacity than 3-dimensional ion traps and, therefore, have an improved trapping efficiency,
which enhances ion signal. In this research, a dynamic fill time was used, which allows the trap
to fill until a certain ion signal is reached as opposed to using a static fill time, which allows the
trap to fill for a specified time, irrespective of ion signal achieved during that time [9, 12]. Spacecharge effects, where ions are ejected from the trap at the wrong time due to overfilling, and
repulsion between ions, are reduced using a dynamic fill time because the trap can be filled
several times with smaller mass ranges [1, 12]. Thus, fewer ions are in the trap at one time,
which reduces repulsion forces between ions and prevents overfilling. Therefore, using a
dynamic fill time can enhance the accuracy of mass spectra and increase the signal [1, 9, 12].
In this research, an enhanced mass spectrum scan was performed, during which all ions
are transmitted through the three quadrupoles (Q 0, Q1, and Q2) without separation [9, 12]. To do
this, RF voltages are applied to these quadrupoles without DC voltages [1, 12]. Then, the ions are
trapped and separated by the LIT [12]. To do this, RF voltages are scanned. For a given RF, only
ions with a specific m/z value are ejected [1, 12]. Therefore, by scanning the voltage, all m/z
values in the desired range can be ejected. The ejected ions then enter the detector, which was a
continuous dynode electron multiplier in this research, similar to that previously described in
Section 2.1, to generate the mass spectrum. Figure 2.9 shows the mass spectrum of salvinorin A
generated using ESI with the pseudomolecular ion, base peak, and major fragments labeled.
Since ESI is a soft ionization method and doesn’t result in substantial fragmentation of
the molecular ion, identification of compounds based on structure is more difficult [1]. To
38
 

1.40E+07

Intensity

Base Peak
373

Molecular Ion + Sodium
455
341
313
295
267
229

0.00E+00
0

Pseudomolecular
2 Molecular Ions + Sodium
Ion
887
433

100 200 300 400 500 600 700 800 900 1000
Mass to Charge (m/z)

Figure 2.9: Mass spectrum of salvinorin A analyzed using electrospray ionization.

39
 

2

2

overcome this, tandem mass spectrometry (MS ) was also performed in this research. In MS ,
either a selected ion or all the ions in the original mass spectrum are fragmented further,
therefore, obtaining another mass spectrum. This provides more information about the
fragmentation patterns of a compound and assists in the determination of an unknown
compound’s structure [1, 10]. This is especially useful with LC analyses because generally
libraries used for identification purposes are not widely available.
2

To perform MS in this research, an enhanced product ion scan was conducted after the
enhanced mass spectrum scan. For the enhanced product ion scan, Q 0 collects the ions [10]. The
second quadrupole (Q1) is set to a specific DC and RF voltage to allow only ions with a specific
m/z value to pass [1, 10]. In this research, the LC-MS/MS was programmed to individually select
the top two abundant ions from the enhanced mass spectrum scan for each compound [9]. After
the ions of a specific compound are selected in Q1, the ions enter the third quadrupole (Q2),
which acts as a collision cell filled with nitrogen gas [1, 10]. The energy of the collision is
transferred into internal energy, which causes the ions to fragment. This collision energy is a
controlled parameter and can vary slightly by using a collision energy spread to allow the
fragmentation of various ions [10]. The ions and their fragments exit the collision cell and are
collected and separated by the LIT, prior to detection [1, 10].

2.2 Data Pretreatment
Chromatographic data collected over a period of time will contain non-chemical sources
of variance, which can include rising chromatographic baselines, electronic noise, retention time
drift, and differences in abundance due to differences in injection volume and detector response
40
 

[13]. Prior to any statistical procedures, data pretreatment should be performed to reduce these
non-chemical sources of variance [13-16]. Once data pretreatment procedures are performed,
patterns in the data due to actual chemical differences become more recognizable. In this
research, four pretreatment procedures were applied to the data: background correction,
smoothing, retention time alignment, and normalization. Each procedure is discussed in more
detail below.

2.2.1 Background Correction
When analyzing samples by GC-MS with high oven temperatures, the stationary phase
can degrade or “bleed” [1, 4]. There are several common bleed ions; however, the identity of
these ions will depend on the type of column used. For example, some common column bleed
ions for a non-polar stationary phase composed of a polysiloxane backbone include m/z 207, 281,
and 355 [1]. Column bleed causes chromatograms to have a rising baseline that is not
reproducible, even for replicate injections of a sample. To minimize this rising baseline,
background correction can be performed [1, 16]. Although there are a number of ways to
perform background correction, in this research, the correction was performed using instrument
software [16]. Using this method, the mass spectrum of the background ions must first be
obtained [16, 17]. This was done by selecting a section of the TIC on the plateau of the rising
baseline that did not contain any sample peaks. The ions in the corresponding mass spectrum are
then subtracted from the full TIC [17]. This will reduce the rising baseline; however, caution is
needed because smaller peaks in the TIC may also be subtracted.
This background correction method is fast, simple, and reproducible. In addition, the

41
 

instrument software is available in most forensic laboratories and, thus, this background
correction method could be easily incorporated into existing laboratory procedures.

2.2.2 Smoothing
Chromatograms contain electronic noise, which is random short-term fluctuations in
detector response [13, 18]. Smoothing algorithms can be applied to reduce this non-chemical
source of variance [13]. While several types of smoothing algorithms are available, a SavitzkyGolay smooth was used in this research [13, 18].
The Savitzky-Golay algorithm fits a polynomial to each section of the chromatogram
using the least squares method [13, 18, 19]. The number of data points in a section of the
chromatogram and order of the polynomial are defined by the user [13]. One factor to consider in
determining the number of data points per section and the order of the polynomial is the number
of data points within a peak. For each section, the polynomial is fit to the data points [13, 18]. To
do this, the algorithm determines the coefficients of the polynomial so that the squared difference
between the y values of the polynomial and the data points is minimal [13, 18, 20]. Using this
polynomial, the best fit data point for the center of the section is then calculated [13, 18– 20].
This is done by substituting the x value of the original center data point into the polynomial to
determine the y value for the new center data point [13, 18]. This new center data point will
replace the original. Then, the section is moved one data point further and the process is repeated
[18]. Although smoothing can reduce noise, chromatograms can be over smoothed, which can
reduce the abundance of compounds [13]. Therefore, a balance must be reached between
reducing noise and minimizing peak reduction.

42
 

2.2.3 Retention Time Alignment
Retention time drift is a common observation in chromatographic data. Variations in
retention time can be due to slight differences in the mobile phase flow rate, mobile phase
composition, oven temperature, and changes in the stationary phase composition due to
degradation [15, 21]. Different algorithms are available to reduce retention time drift. In this
research, a correlation optimized warping (COW) algorithm and a peak matching algorithm were
used.
For both algorithms, the sample TICs are aligned to a target TIC, which should include as
many peaks as possible from the TICs to be aligned [22]. The target TIC can be selected at
random from the data set, generated as the average of all TICs in the data set, or prepared from a
sample containing all analyte compounds that is then analyzed to generate the target [14, 21, 22].
For the COW algorithm, the target and the sample TICs are each divided into segments
[14, 15, 21]. Two parameters can be adjusted: segment size and warp size [14, 21, 22]. Segment
size is the number of data points per segment, while warp size is the maximum number of data
points that can be added or subtracted from a segment in order to adjust the alignment of a peak
[14, 15, 21, 22]. Each segment in the sample TIC is separately aligned to the target starting at the
end of the TIC and the number of data points defined by the warp is added or subtracted [14].
For example, if the warp size is set to 3, then 0, 1, 2, or 3 data points can be added or subtracted
from a segment in the sample TIC. Local correlation coefficients are calculated between the
segment in the sample TICs and the target TIC for each warp [14, 21]. The warp giving the
highest local correlation coefficient is used, although all coefficients for all warp sizes in that
segment are also stored. The algorithm moves to the next segment to repeat the procedure. After
aligning each segment, a global correlation coefficient is calculated, which considers all

43
 

combinations of the local correlation coefficients for each segment [14]. The combination of
warps that give the highest global correlation coefficient is considered the optimal alignment
[21].
For the peak matching algorithm, noise and peak identification thresholds are first
defined [23]. The noise is defined as the standard deviation of a range of data points toward the
beginning or end of the TICs. In this research, the noise was defined as the standard deviation of
88 data points towards the end of the TICs. The peak identification threshold, which is a
threshold used to define a peak, is calculated by multiplying the noise by a constant. In this
research, two times the noise was used. Any signal above this value was identified as a peak.
The algorithm then finds the peaks in the target TIC [23]. This is done by calculating the
first derivative at each data point in the TIC. The leading edge of a peak is identified when the
difference in abundance between consecutive data points is greater than the peak identification
threshold. The algorithm then searches for the zero crossing in the first derivative, which
represents the peak maximum, while the next zero crossing represents the tailing edge of that
peak. The retention time corresponding to the zero crossing is interpolated. This retention time,
representing the apex of the peak, is added to a list containing retention times of all peaks
identified in the target. The algorithm is then applied to each sample TIC, which generates tables
of peaks identified in each.
Once peaks are identified, the retention times of identified peaks in the sample TICs are
compared to those identified in the target TIC [23]. If a peak is present in the target and sample
TIC with retention times within a defined window size, which is a specific number of data
points, then the peaks are considered a “match.” The algorithm then interpolates the retention

44
 

time axis to align the peak maxima. This procedure is repeated for each sample TIC in order to
align the data.

2.2.4 Normalization
Slight variation in injection volume and detector response can lead to differences in
abundance of the same compounds in replicate injections [13]. To account for these differences,
normalization can be performed prior to data analysis. While several types of normalization are
available, total area and internal standard normalization were used in this research. For total area
normalization, the total area of each TIC is first calculated [13, 24]. Then, the average total area
of all TICs in the data set is calculated. For each TIC, the abundance at each retention time is
divided by the total area of that TIC, and then multiplied by the average area of all TICs in the
data set.
For internal standard normalization, the peak abundance of the internal standard in each
TIC is obtained [13]. Then, the average peak abundance of the internal standard in all TICs in the
data set is calculated. For each TIC, the abundance at each retention time is divided by the
internal standard abundance for that TIC, and then multiplied by the average internal standard
abundance of all TICs in the data set.

2.3 Data Analysis
2.3.1 Pearson Product Moment Correlation Coefficients
Pearson product moment correlation (PPMC) coefficients are used to make pair-wise
comparisons between two samples [25]. The coefficients are calculated using equation 2.1,
∑
∑
̅

̅
∑

45
 

/                                             (2.1) 

where r is the correlation coefficient, i and i represent the individual data points for two
different samples being compared, and ̅  and

are the average data from the samples. In this

research, PPMC coefficients were calculated between chromatograms of the Salvia species.
Therefore, i and i represent the abundance of data point i in each chromatogram (x and y),
while ̅ and

are the average abundance of all data points in each chromatogram. The coefficient

ranges from 0 to ±1, with the sign indicating a positive or negative correlation. A positive
correlation means the data points in both chromatograms increase or decrease at the same time,
while a negative correlation means the data points of one chromatogram increase while the data
points of the other decrease [26]. A coefficient of 0 indicates no correlation between the two
samples being compared, while a coefficient of 1 indicates a perfect correlation. Coefficients
between ±0.8 to ±1 indicate strong correlation, ±0.5 to ±0.79 indicate moderate correlation, and
less than ±0.5 indicate weak correlation [27]. 

2.3.2 Principal Components Analysis
Principal components analysis (PCA) is an unsupervised, multivariate statistical
procedure used to reduce the dimensionality of the data set [13, 16, 26, 28, 29]. Therefore, PCA
simplifies the data and reveals underlying patterns in the data.
To perform PCA, the data are first mean-centered to set the average of the data set to zero
[16, 26, 29]. To do this, the average of the data points across all samples is calculated and then
the average is subtracted from each data point in each sample [26]. Next, the covariance between
each pair of samples is calculated using equation 2.2,

	

,

∑
̅

46
 

(2.2)

where X and Y are different samples, ̅ and

are the averaged data from the samples, and n is

the number of samples [26]. The covariance for all pair-wise comparisons is displayed as a
covariance matrix that is symmetrical around the diagonal [26, 30].
The eigenvectors and eigenvalues of the covariance matrix are then calculated [26, 29].
An eigenvector is a unit vector that can be multiplied by a matrix and results in a product that is a
multiple of the unit vector [26]. The multiple of the eigenvector is called the eigenvalue. For
PCA, each eigenvector represents a principal component, which describes the variance among
the samples [16, 26]. Each successive eigenvector is perpendicular to the next [26]. The first
principal component (PC1) describes the most variance in the data set, while the second principal
component (PC2) describes the next greatest amount of variance. The amount of variance
described is given by the eigenvalue for that eigenvector, in proportion to all other eigenvalues
[13, 16, 26].
The outputs from PCA are scores and loadings plots [16]. The score for a sample on PC1
is calculated by multiplying the PC1 eigenvector by the mean-centered data for that sample, then
summing the products [26]. This procedure can be repeated using the PC2 eigenvector to
calculate the score for the sample on PC2. The scores of each sample, calculated in a similar
manner, are then plotted to generate the scores plot, which typically shows PC1 on the x-axis and
PC2 on the y-axis [16, 26, 28, 30]. The positioning of the samples on the scores plot can be used
to associate and discriminate samples based on their relative location [16, 30]. Samples that are
closely clustered in the scores plot are chemically similar, while those samples separated in the
scores plot are chemically different [13, 16, 28].
The loading plots can be used to identify the most variable chemical compounds among
the samples for each principal component [16]. The loadings plots for chromatographic data can

47
 

be generated by graphing the eigenvector for each principal component versus retention time.
Compounds can either load positively or negatively in these plots and the height of each
compound in the plots corresponds to its contribution to the positioning of the samples in the
scores plot [13, 16, 28]. Since the loadings plots consist of the eigenvector for each PC plotted
against retention time, it is possible to identify the compounds in the samples that are
contributing to the variance described by each PC [16].

2.3.3 Student’s t-tests
Student’s t-tests are hypothesis tests used to determine if there is a statistically significant
difference between two sample populations [25]. The null hypothesis (H0) states that the two
sample populations are equal, while the alternative hypothesis (H1) states they are significantly
different. In this research, Student’s t-tests were calculated on the mean scores of each species to
determine if Salvia divinorum could be statistically differentiated from four other Salvia species.
All variances were assumed to unequal. Therefore, the Student’s t-test for unequal variances was
used (equation 2.3),
̅

	

where

̅

and

̅

̅

(2.3)

are the average of sample populations 1 and 2, respectively, s1 and s2 are the

standard deviation of sample populations 1 and 2, respectively, and n1 and n2 are the number of
measurements in sample populations 1 and 2, respectively [25]. For unequal variance, the
degrees of freedom (υ) is calculated using equation 2.4,

48
 

	

(2.4)

The degrees of freedom is used to determine the tcritical value from statistical tables at the desired
confidence level [31]. When determining tcritical, it must be decided if a one-tailed or two-tailed
t-test is necessary. A one-tailed t-test is used to determine whether one sample population is
significantly greater or significantly less than the other [25]. A two-tailed t-test is used to
determine whether the sample populations are significantly different, irrespective of one sample
population being greater or less than the other. Tcritical from the table for the desired tailed t-test,
confidence level, and degrees of freedom is then compared to tcalculated from equation 2.3 [25,
31]. If tcalculated is less than tcritical, then H0 is accepted and the two sample populations are
equal [25]. If tcalculated is greater than tcritical, then H0 is rejected, H1 is accepted, and the two
sample populations are significantly different from one another.

49
 

References

50
 

References
[1] Watson JT, Sparkman OD. Introduction to mass spectrometry: instrumentation, applications
and strategies for data interpretation. 4th ed. West Sussex, England: John Wiley & Sons Ltd,
2007.
[2] Skoog DA, Holler FJ, Crouch SR. Principles of instrumental analysis. 6th ed. Belmont,
California: Thomson Higher Education, 2007.
[3] McMaster M, McMaster C. GC/MS: A practical user’s guide. New York: Wiley-VCH, 1998.
[4] Heftmann E, editor. Chromatography part a: fundamentals and techniques. 6th ed.
Amsterdam, The Netherlands: Elsevier B. V., 2004.
[5] Willard HH, Merritt Jr. LL, Dean JA, Settle Jr. FA. Instrumental methods of analysis. 7th ed.
Belmont, California: Wadsworth Publishing Company, 1988.
[6] Reid GE. Ionization methods: How are ions formed? Class notes from CEM 832. Spring
2010.
[7] Tsujikawa K, Kuwayama K, Miyaguchi H, Kanamori T, Iwata YT, Yoshida T, Inoue H.
Determination of salvinorin A and salvinorin B in Salvia divinorum-related products circulated
in Japan. Forensic Sci Int 2008; 180:105–9.
[8] Grilli M, Neri E, Zappettini S, Massa F, Bisio A, Romussi G, et al. Salvinorin A exerts
opposite presynaptic controls on neurotransmitter exocytosis from mouse brain nerve terminals.
Neuropharmacology 2009; 57:523–30.
[9] Jones AD. LC-MS training at Michigan State University Mass Spectrometry Facility. 2010.
[10] Jones AD. Background and theory of LC/MS/MS on QTRAP mass spectrometers.
Powerpoint from Michigan State University Mass Spectrometry Training Facility.
[11] Gabelica V, De Pauw E. Internal energy and fragmentation of ions produced in electrospray
sources. Mass Spectrometry Reviews 2005; 24:566–87.
[12] Jones AD. Personal Communication. December 2011.
[13] Blackledge RD, editor. Forensic analysis on the cutting edge: new methods for trace
evidence analysis. New Jersey: John Wiley & Sons, 2007.
[14] Tomasi G, Berg FV, Andersson C. Correlation optimized warping and dynamic time
warping as preprocessing methods for chromatographic data. Journal of Chemometrics 2004;
18:231–41.

51
 

[15] Malmquist G, Danielsson R. Alignment of chromatographic profiles for principal
component analysis: a prerequisite for fingerprinting methods. Journal of Chromatography A
1994; 687:71–88.
[16] Brereton RG. Applied chemometrics for scientists. West Sussex, England: John Wiley &
Sons, 2007.
[17] MSD ChemStation Software version E.01.01.335 Agilent Technologies, Inc, 1987-2007.
[18] Savitzky A, Golay MJE. Smoothing and differentiation of data by simplified least squares
procedures. Analytical Chemistry 1964; 36:1627–39.
[19] Gorry PA. General least-squares smoothing and differentiation by the convolution
(Savitzky-Golay) method. Analytical Chemistry 1990; 62:570–3.
[20] Madden HH. Comments on the Savitzky-Golay convolution method for least-squares fit
smoothing and differentiation of digital data. Analytical Chemistry 1978; 50:1383–6.
[21] Nielsen NPV, Carstensen JM, Smedsgaard J. Aligning of single and multiple wavelength
chromatographic profiles for chemometric data analysis using correlation optimized warping.
Journal of Chromatography A 1998; 805:17–35.
[22] Skov T, van den Berg F, Tomasi G, Bro R. Automated alignment of chromatographic data.
Journal of Chemometrics 2006; 20:484–97.
[23] Johnson KJ, Wright BW, Jarman KH, Synovec RE. High-speed peak matching algorithm
for retention time alignment of gas chromatographic data for chemometric analysis. Journal of
Chromatography A 2003; 996:141–55.
[24] Rietjens M. Reduction of error propagation due to normalization: effect of error propagation
and closure on spurious correlations. Analytica Chimica Acta 1995; 316:205–15.
[25] Miller JN, Miller JC. Statistics and chemometrics for analytical chemistry. 4th ed. Harlow,
England: Prentice-Hall, 2000.
[26] Smith LI. A tutorial on principal components analysis. 2002. http://www.cs.otago.ac.nz/
cosc453/student_tutorials/principal_components.pdf
[27] Devore JL. Probability and statistics for engineering and the science. 4th ed. Duxbury Press,
Belmont, CA, 1995.
[28] Bodnar Willard MA. Forensic analysis of Salvia divinorum and related Salvia species using
chemometric procedures. MS Thesis (2010).
[29] Shlens J. A tutorial on principal component analysis: version 2. Systems Neurobiology
Laboratory, Salk Insitute for Biological Studies. La Jolla, California 2005. http://www.snl.
salk.edu/~shlens/pca.pdf
52
 

[30] Lee JLS, Gilmore IS, Seah MP. Proposed terminology for multivariate analysis in surface
chemical analysis – vocabulary – part 1: general terms and terms for the spectroscopies. National
Physical Laboratory. Teddington, Middlesex, UK. 2008. http://www.npl.co.uk/upload/pdf/
multivariate_analysis.pdf
[31] Adam C. Essential mathematics and statistics for forensic science. West Sussex, England:
John Wiley & Sons Ltd, 2010.

53
 

Chapter 3 Association and Discrimination of Salvia divinorum from Related Salvia
Species Based on Chemical Fingerprints Generated Using Gas Chromatography-Mass
Spectrometry
3.1 Introduction
This chapter describes the association and discrimination of S. divinorum from related
Salvia species based on their volatile chemical compounds. S. divinorum and four other
Salvia species were extracted in dichloromethane and extracts were analyzed by gas
chromatography-mass spectrometry (GC-MS). The resulting total ion chromatograms (TICs)
were background corrected, truncated, smoothed, retention time aligned, and normalized.
Mean Pearson product moment correlation (PPMC) coefficients were then used to evaluate
the precision of the extraction and analysis procedures, and to evaluate the association of S.
divinorum extracts and the discrimination of S. divinorum from the other four Salvia species.
Principal components analysis (PCA) was used to assess the discrimination of S. divinorum
from the other four Salvia species in the scores plot. The PCA loadings plots were used to
identify the compounds contributing most to the variance among the species. Finally,
Student’s t-tests were used to statistically evaluate the association and discrimination in the
PCA scores plot.

3.2 Material and Methods
3.2.1 Extraction of Salvia Species
Leaves of S. divinorum were obtained from the research greenhouse at Michigan State
University (MSU). S. guaranitica, S. nemorosa, and S. splendens were purchased from
Christian’s greenhouse in Haslett, Michigan. S. officinalis was purchased from Van Atta’s
Greenhouse and Flower Shop in Haslett, Michigan. Plants were transported to the MSU research

54
 

greenhouse, where they were transplanted as needed. Leaves were collected from each species,
briefly rinsed with water to remove any dirt, and placed in a salad spinner (OXO, Chambersburg,
PA) to remove the excess water. All leaves were dried in the same manner, one species at a time,
with a food dehydrator (NESCO American Harvest model ED-75PR, Two Rivers, WI) at 35 °C.
All food dehydrator shelves and liners were cleaned between drying each species. S. guaranitica,
S. nemorosa, and S. splendens were dried for 12 hours, while S. divinorum and S. officinalis were
dried for 16 hours and 36 hours, respectively. Leaves were then stored in sealed bags at 1 °C
until analysis.
All glassware used for the extraction was acid-washed following standard procedures.
Approximately 0.04 g dried leaves were placed in a 100 mL round bottom flask and 13 mL
dichloromethane (J. T. Baker, Phillipsburg, NJ) were added. The flask was placed on a rotary
agitator (Büchi Rotavapor-R, Flawil, Switzerland) for 16 hours at medium speed. The extract
was then filtered using the experimental set-up in Figure 3.1. The watch glass was placed on top
of the Büchner funnel containing the nylon mesh filter (Small Parts Inc., Miami Lakes, FL) to
generate a stronger vacuum. The rotary flask was then rinsed with 2 mL dichloromethane (J. T.
Baker), the rinsings were filtered, and the entire extract was evaporated to dryness using nitrogen
gas (Airgas, Independence, OH) and a water circulating bath (NESLAB endocal RTE-9B,
Newington, NH) at 35 °C. This procedure was repeated in triplicate for each of the Salvia
species, using fresh leaf material for each extraction. All extracts were stored at 2 °C until
analysis.

55
 

Watch glass
Büchner funnel
with nylon
mesh filter

Rubber
tubing
Aspirator
Water

Vacuum flasks
Figure 3.1: Schematic diagram of the filtration system.

56
 

3.2.2 Gas Chromatography-Mass Spectrometry Analysis
All extracts were reconstituted in 0.58 mL dichloromethane (J. T. Baker) and
progesterone (Sigma Chemical Company, St. Louis, MO) was added as an internal standard
(0.00119 M). Extracts were shaken, vortexed, and each was divided into separate autosampler
vials (Restek, Bellefonte, PA) with inserts. This allowed for replicate analysis of each extract
while minimizing evaporative loss due to puncturing the septum of the vial. Extracts were
analyzed using an Agilent 6890N Network GC system (Agilent Technologies, Santa Clara, CA)
with a 5973 inert mass selective detector (Agilent Technologies), equipped with ChemStation©
software (version E 01.01.335). Extracts were injected via an Agilent 7683 Series autosampler
(Agilent Technologies) using a 1 µL injection volume and a split ratio of 50:1. A J&W DB-5MS
column (30 m x 0.25 mm x 0.25 µm, Agilent Technologies) was used. The inlet and mass
transfer line temperatures were maintained at 300 °C. The carrier gas was ultrahigh purity helium
(Airgas, Independence, OH), with a nominal flow rate of 1.0 mL/min. The oven temperature
program was as follows: 80 °C for 2 minutes, then 10 °C/min to 340 °C, hold for 8 minutes.
Electron ionization (70 eV) was used with a source temperature of 230 °C. The mass analyzer
was a single quadrupole and was operated in full scan mode over the range m/z 50–550. Spectra
of compounds were then collected using ChemStation© and provisional identification of the
major compounds in each extract was made by comparison with the NIST MS Search 2.0 library
(The NIST Mass Spectral Search Program, Version 2.0 d, Gaithersburg, MD) and various
literature sources [1–7].

57
 

3.2.3 Data Pretreatment
Each extract was analyzed in triplicate, which resulted in a total of nine TICs for each
Salvia species. All TICs were pretreated prior to data analysis. As a result of using high oven
temperatures, chromatograms had a rising baseline due to column bleed. This would cause
problems in subsequent PCA because differences in the baseline abundance would be identified
as the major source of variance among the extracts. Therefore, this rising baseline was reduced
by subtracting the background using the ChemStation© software. To do this, the mass spectrum
at retention time 35 minutes was selected to represent the background spectrum. This was
repeated for each TIC. The selected background was then subtracted from each TIC. To further
reduce the rising baseline, TICs were truncated at approximately 30.6 minutes since no
compounds were identified past this time in any of the species. Truncated TICs were then
smoothed to reduce electronic noise. A Savitzky-Golay smooth was performed in OriginPro 7.5
(Version 7.5, OriginLab, Guangzhou, China) using a second order polynomial and a window size
of 5 points.
To correct for retention time drift due to instrumental variability, TICs were aligned to an
average target chromatogram, which was created using one TIC of each Salvia species. A
random number generator in Microsoft Excel (Versions 2007 and 2010, Microsoft Corporation,
Redmond, WA) was used to select the TICs to create the target. The five TICs were imported
into Microsoft Excel and the average abundance at each retention time was calculated to generate
the target chromatogram. Total ion chromatograms were firstly aligned to the target
chromatogram using a correlation optimized warping (COW) algorithm (LineUp, Version 3.0,
InfoMetrix, Inc., Bothell, WA) with a warp of 5 data points and segment size of 65 data points.
Then, the TICs were aligned again, using a peak match algorithm available in the literature,

58
 

which was performed in Matlab (Version 7.11.0.584 R2010b, The Mathworks, Natick, MA) [8].
For this algorithm, the noise was defined as the standard deviation of 88 data points in the
retention time range 29.8 to 30.3 minutes, a peak was defined as two times the standard deviation
of the noise, and a window size of 2 was used.
Following alignment, the TICs were normalized in a two-step procedure in Microsoft
Excel. Firstly, TICs of each species were subjected to total area normalization. To do this, the
area under each TIC was determined, as was the average area of the TICs of that species. Then,
the abundance at each retention time in a TIC was divided by the area of that TIC and multiplied
by the average area of the TICs of the same species. Following total area normalization, the TICs
were subjected to internal standard normalization. To do this, the abundance at each retention
time was divided by the abundance of the internal standard in the TIC, and then multiplied by the
average abundance of the internal standard in all TICs of that species.

3.2.4 Data Analysis
PPMC coefficients were calculated in Matlab on pretreated TICs for all pair-wise
comparisons of replicate injections of the same extract and replicates of each species to assess
the precision of the extraction and analysis procedures. Principal components analysis was then
performed on the pretreated data using Matlab. Scores and loadings plots were generated in
Microsoft Excel. To statistically assess discrimination of S. divinorum from the four Salvia
species, Student’s t-tests were performed in Microsoft Excel on the mean scores of each species
on both principal components (PCs) 1 and 2 at various confidence levels. To further assess
association and discrimination of S. divinorum from the four Salvia species, PPMC coefficients
were also calculated on TICs for all pair-wise comparisons of the species.

59
 

3.3 Results and Discussion
3.3.1 Provisional Identification of Compounds Present in Salvia Species
Representative TICs of the five Salvia species are shown in Figures 3.2–3.6. Provisional
identification of compounds present in each species, along with molecular masses and major
fragment ions, are listed in Appendix A. S. nemorosa and S. officinialis belong to Salvia clade I,
while S. divinorum, S. guaranitica, and S. splendens belong to Salvia clade II [2, 9, 10]. Species
in different clades contained common compounds; therefore, classification of the species into
clades based on the volatile compounds was difficult. For example, all five Salvia species
contained hentriacontane, tetratriacontane, and sitosterol [4]. S. guaranitica, S. nemorosa, S.
officinalis, and S. splendens all contained vitamin E. Squalene, nonacosane, and heptacosane
were also present in two or more species.
While there were many common compounds, certain compounds were present in only
one species and, hence, could be used to distinguish species. For example, S. divinorum was the
only species to contain salvinorins A, B, D, an acetate compound, and pentatriacontane [1, 2]. S.
officinalis contained many early eluting compounds, such as thujone, camphor, and epimanool,
while the other species had few or no early eluting compounds [11]. S. guaranitica was the only
species to contain a phthalide. S. nemorosa was the only species to contain β-caryophyllene and
S. splendens was the only species to contain cardenolide. Although these five Salvia species
could be differentiated based on visual assessment of the TICs, there are over 900 Salvia species,
some of which may be more difficult to distinguish from S. divinorum based only on visual
assessment of TICs. Therefore, statistics were applied to this research as a proof-of-concept
study.

60
 

1.30E+06

3

Abundance

8

12

0.00E+00
3

4-7

Retention Time (minutes)

9-11
12
36

Figure 3.2: Total ion chromatogram of Salvia divinorum extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry. Provisional identifications based on
library searches and literature sources are as follows: 1=heptacosane, 2=squalene,
3=progesterone, 4=nonacosane, 5=an acetate, 6=hentriacontane, 7=salvinorin B, 8=salvinorin A,
9=tetratriacontane, 10=salvinorin D, 11=sitosterol, and 12=pentatriacontane [1, 2]. See Appendix
A, Table A.1 for more detailed information.

61
 

1.30E+06

Abundance

3

15,16
19

13 14
0.00E+00
3

2

20
17
11
22
18 6 9 23 21

Retention Time (minutes)

36

Figure 3.3: Total ion chromatogram of Salvia guaranitica extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry. Provisional identifications based on
library searches and literature sources are as follows: 2= squalene, 3=progesterone, 6=
hentriacontane, 9=tetratriacontane, 11=sitosterol, 13=phthalide, 14=β-cubebene, 15= an ester,
16= an alkene, 17=vitamin E, 18=dione, 19=dione, 20=nitrolepidine, 21=benzenesulfonamide,
22=β-amyrin, and 23=α-amyrin [5, 6]. See Appendix A, Table A.2 for more detailed
information.

62
 

1.30E+06

Abundance

3

9

24 14

0.00E+00
3

25

26

27, 28
17
6 11
29
4

Retention Time (minutes)

Figure 3.4: Total ion chromatogram of Salvia nemorosa extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry. Provisional identifications based on
library searches and literature sources are as follows: 3=progesterone, 4=nonacosane,
6=hentriacontane, 9=tetratriacontane, 11=sitosterol, 14=β-cubebene, 17=vitamin E, 24=βcaryophyllene, 25=alkene, 26=amide, 27=phenyl, 28=11-n-decyltetracosane, and
29=hextriacontane [4]. See Appendix A, Table A.3 for more detailed information.

63
 

36

1.30E+06

35

3

37

Abundancle

9
45
46
34
32,
33
30
31

0.00E+00
3

36

40

17 50

42

38
39 41

43

6
47
49
481,2 4

44 25
Retention Time (minutes)

11

51, 52, 23, 29
53

36

Figure 3.5: Total ion chromatogram of Salvia officinalis extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry. Provisional identifications based on
library searches and literature sources are as follows: 1=heptacosane, 2=squalene,
3=progesterone, 4=nonacosane, 6=hentriacontane, 9=tetratriacontane, 11=sitosterol, 17=vitamin
E, 23=α-amyrin, 25=alkene, 29=hexatriacontane, 30=α-pinene, 31=camphene, 32=β-pinene, 33=
β-myrcene, 34=eucalyptol, 35=thujone, 36=thujone, 37=camphor, 38=borneol, 39=bornyl
acetate and/or isoborynyl acetate, 40=bicyclic compound, 41=a cyclopropane compound, 42=αcaryophyllene, 43=epiglobulol, 44=diene, 45=epimanool, 46=isocarnosol, 47=phenyl, 48=dione,
49=squalane, 50=triacontane, 51=dotriacontane, 52=9-n-octylhexacosane, and 53=oleanolic acid
[3]. See Appendix A, Table A.4 for more detailed information.

64
 

1.30E+06

3
2

Abundance

54

11
17 9
55, 23
25

0.00E+00
3

6

29 53

4

Retention Time (minutes)

36

Figure 3.6: Total ion chromatogram of Salvia splendens extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry. Provisional identifications based on
library searches and literature sources are as follows: 2=squalene, 3=progesterone,
4=nonacosane, 6=hentriacontane, 9=tetratriacontane, 11=sitosterol, 17=vitamin E, 23=α-amyrin,
25=alkene, 29=hextriacontane, 53=oleanolic acid, 54=8-episalviarin, and 55=cardenolide [7].
See Appendix A, Table A.5 for more detailed information.

65
 

3.3.2 Investigation of Extraction and Analysis Precision Using Pearson Product Moment
Correlation Coefficients
The mean, standard deviation, and range of PPMC coefficients were calculated for
replicate injections (n=3) of each extract to demonstrate precision in the instrumental analysis
procedure. However, since data pretreatment was performed prior to calculating PPMC
coefficients, precision of these procedures was also reflected in the coefficients.
Ideally, replicates are expected to have a PPMC coefficient of 1.00. With the exception of
one extract of S. nemorosa, all mean PPMC coefficients for injection replicates and the ranges of
coefficients were greater than 0.9, which indicated strong correlation (Table 3.6). The lower
mean PPMC coefficient (0.8583 ± 0.1202) and the range of coefficients (0.7366 – 0.9769) for
one extract of S. nemorosa was due to retention time misalignment of the internal standard peak
among replicates. All standard deviations were relatively low (≤ 0.06) except for the aforementioned extract of S. nemorosa. Thus, overall, there was acceptable precision in the
instrumental analysis and data pretreatment procedures.
To demonstrate precision in the extraction procedure, in addition to the instrumental
analysis and data pretreatment procedures, the mean, standard deviation, and range of PPMC
coefficients were also calculated for all extracts (n=36) of each species. The mean PPMC
coefficients were greater than 0.9, which indicated strong correlation between extracts of the
same species (Table 3.7). Lower coefficients were expected since different leaf material was
used for each extraction. The ranges of coefficients were greater than 0.9 with the exception of S.
nemorosa, which had a broader range of coefficients (0.7366 – 0.9981) due to the retention time
misalignment discussed previously. All standard deviation were relatively low; however, S.
nemorosa had a higher standard deviation (0.0563) than the replicates of the other species, again

66
 

Table 3.6: Mean, standard deviation, and range of Pearson product moment correlation (PPMC) coefficients for replicate injections of
the same extract
Mean PPMC coefficient ± standard deviation
Range of PPMC coefficients
Species

Extract 1 (n=3)

Extract 2 (n=3)

Extract 3 (n=3)

Extract 1 (n=3)

Extract 2 (n=3)

Extract 3 (n=3)

S. divinorum

0.9550 ± 0.0304

0.9886 ± 0.0071

0.9953 ± 0.0033

0.9279 – 0.9879

0.9830 – 0.9966

0.9922 – 0.9988

S. guaranitica

0.9968 ± 0.0013

0.9782 ± 0.0176

0.9872 ± 0.0102

0.9959 – 0.9983

0.9659 – 0.9985

0.9767 – 0.9971

S. nemorosa

0.8583 ± 0.1202

0.9958 ± 0.0018

0.9890 ± 0.0062

0.7366 – 0.9769

0.9937 – 0.9969

0.9819 – 0.9934

S. officinalis

0.9766 ± 0.0060

0.9869 ± 0.0053

0.9881 ± 0.0035

0.9728 – 0.9834

0.9811 – 0.9912

0.9853 – 0.9920

S. splendens

0.9869 ± 0.0097

0.9844 ± 0.0113

0.9873 ± 0.0084

0.9770 – 0.9962

0.9715 – 0.9930

0.9776 – 0.9926

67
 

Table 3.7: Mean, standard deviation, and range of Pearson product moment correlation (PPMC)
coefficients for all extracts for each species
Species
Mean PPMC coefficient ± standard Range of PPMC coefficients
(n=36)
deviation (n=36)
S. divinorum

0.9696 ± 0.0243

0.9156 – 0.9988

S. guaranitica

0.9828 ± 0.0163

0.9368 – 0.9985

S. nemorosa

0.9575 ± 0.0563

0.7366 – 0.9981

S. officinalis

0.9703 ± 0.0152

0.9370 – 0.9920

S. splendens

0.9874 ± 0.0086

0.9670 – 0.9971

68
 

due to the retention time misalignment. Overall, the high PPMC coefficients indicated acceptable
precision in the extraction, instrumental analysis, and data pretreatment procedures.

3.3.3 Association and Discrimination of S. divinorum from Other Salvia Species
The scores plot (Figure 3.8) of the first and second principal components (PC1 and PC2,
respectively) accounted for 77% of the total variance among the five Salvia species. S. officinalis
was the only species positioned positively on PC1. On PC2, S. splendens was positioned
positively, S. divinorum was positioned negatively, and S. guaranitica, S. nemorosa, and S.
officinalis were positioned around zero. Replicates of S. splendens, S. divinorum, and S.
guaranitica were clustered closely; however, replicates of S. officinalis were spread in PC1 and
PC2, while replicates of S. nemorosa were spread in PC2.
The positioning of each species, and the spread observed among replicates, can be
explained with reference to the loadings plots for PC1 and PC2, which contain the compounds
contributing most to the variance among the five species (Figure 3.9).
The majority of the compounds weighted positively on PC1 in the loadings plot (e.g., αpinene, camphene, eucalyptol, thujone, camphor, epiglobulol, epimanool, isocarnosol) were
compounds present in S. officinalis (Figure 3.9A). While some compounds present in S.
officinalis were weighted negatively on this PC, their contribution was minimal in comparison to
the positively- weighted compounds. Therefore, S. officinalis was positioned positively on PC1
in the scores plot. Spread observed among replicates of S. officinalis on PC1 in the scores plot
was primarily due to differences in the abundance of the more volatile compounds, thujone,
camphor, and epimanool, which were weighted positively on this PC.

69
 

PC2 (28%)

7.00E+06

0.00E+00

-7.00E+06

7.00E+06

Replicate 1A

S. divinorum
S. officinalis

-7.00E+06
PC1 (49%)
S. guaranitica
S. nemorosa
S. splendens

Figure 3.8: Scores plot of principal component 1 versus principal component 2 showing
discrimination of S. divinorum from the other four Salvia species based on chemical profiles
obtained by gas chromatography-mass spectrometry.

70
 

0.35

A

Loadings on PC1

d

a–e

b

IS
a–e
b, c

e a, b
b b

a, b, d, e
-0.10
3

Retention Time (minutes)

a–e

31

0.6

B

Loadings on PC2

IS

e
b –e
a–e

-0.1
3

b c, d
d
Retention Time (minutes)

a– e

a, b
31

Figure 3.9: Loading plots for (A) principal component 1 (PC1) and (B) principal component 2
(PC2) based on principal components analysis of the five Salvia species analyzed by gas
chromatography-mass spectrometry. Selected compounds are labeled based on species as
follows: a=Salvia divinorum, b=Salvia guaranitica, c=Salvia nemorosa, d=Salvia officinialis,
e=Salvia splendens. IS=internal standard. See Appendix A for provisional identifications based
on library searches and literature sources.
71
 

S. officinalis was positioned around zero on PC2 in the scores plot: two replicates of this
species were positioned slightly positively, while the other seven replicates were positioned
slightly negatively. This difference in positioning was due to differences in abundance of
compounds such as hentriacontane, vitamin E, sitosterol, which were weighted positively on
PC2, and thujone and camphor, which were weighted negatively on PC2. For example, the
replicate of S. officinialis positioned most positively on PC2 in the scores plot contained the
highest abundance of hentriacontane, vitamin E, and sitosterol, and the lowest abundance of
thujone and camphor, while the replicate positioned most negatively on PC2 contained the
lowest abundance of hentriacontane, vitamin E, and sitosterol.
In addition to the above problems, a retention time misalignment of the internal standard
peak in the TICs also contributed to the positioning and spread among S. officinalis replicates on
PC2 of the scores plot. Figure 3.10 shows this misalignment among the TICs of nine replicates of
S. officinalis and one replicate of S. nemorosa. The effect of the misalignment was observed
when the TICs were mean centered, which was necessary for PCA. Figure 3.11A shows the
mean-centered data for one replicate of S. officinalis. Only one peak should be apparent for each
compound; however, due to the misalignment of the internal standard previously mentioned, two
peaks were observed corresponding to the internal standard. One peak had a retention time (tR)
of 24.44 min, while the other peak was at tR 24.46 min.
In order to mean center a data set, the average abundance at each retention time for all
sample TICs was calculated. This average was then subtracted from the abundance at each
retention time in each TIC to generate the mean-centered data. The replicate of S. officinalis
illustrated in Figure 3.11 contained the highest abundance of internal standard among the
replicates of this species at 24.44 min and, therefore, when the average abundance of all TICs
72
 

Abundance

6.00E+06

0.00E+00
24.41

24.42

24.43 24.44 24.45 24.46
Retention Time (minutes)

24.47

24.48

Figure 3.10: Magnified version of nine total ion chromatograms of S. officinalis (solid line) and
one total ion chromatogram of S. nemorosa (dashed line) showing the misalignment of the
internal standard peak.

73
 

1.40E+06

Abundance

A
IS

0.00E+00
IS

-1.40E+06
3

Retention Time (minutes)

31

2.50E+05

B

Loadings

IS

0.00E+00

IS

-2.50E+05
3

Retention Time (minutes)

31

Figure 3.11: (A) Mean-centered data of one replicate of S. officinalis and (B) loadings for the
same replicate of S. officinalis on principal component 2 based on chemical profiles obtained by
gas chromatography-mass spectrometry with progesterone added as the internal standard (IS).

74
 

was subtracted, the result was a positive contribution at this retention time in the mean-centered
data. However, this same replicate had the lowest abundance of internal standard among the
replicates at 24.46 min in the TIC. Therefore, when the average abundance of all TICs at this
retention time was subtracted, the result was a negative contribution at this retention time in the
mean-centered data.
The mean-centered data were then multiplied by the eigenvector for PC2, which
generated the loadings, and every data point was summed to generate the score for the sample on
PC2. The eigenvector was positive at these retention times (Figure 3.9B) such that, when
multiplied by the mean-centered data, the product was positive at 24.44 min and negative at
24.46 min (Figure 3.10B). The internal standard was weighted heavily. For this replicate, the
positive loading of the internal standard outweighed the negative loading such that this replicate
was positioned positively on PC2 in the scores plot. All replicates of S. officinalis contained two
internal standard peaks in their PC2 loadings. Therefore, a combination of differences in
abundance of various compounds including the internal standard, as well as a misalignment of
the internal standard, contributed to the positioning of S. officinalis replicates on PC2.
S. divinorum was positioned negatively on both PC1 and PC2 in the scores plot. From the
corresponding loadings plots, salvinorin A, the main compound in S. divinorum, was weighted
negatively on both PCs, which resulted in the observed positioning in the scores plot.
S. guaranitica was positioned negatively on PC1 in the scores plot. Some compounds
present in S. guaranitica (i.e. phthalide, hentriacontane, tetratriacontane, and β-amyrin) were
weighted positively on PC1 (Figure 3.9A). However, many more compounds in this species
(e.g., β-cubebene, squalene, sitosterol, and α-amyrin) were weighted negatively. Overall, S.
guaranitica was positioned negatively on PC1 in the scores plot.

75
 

S. guaranitica was positioned around zero on PC2. In this case, there were a similar
number of compounds weighted positively as weighted negatively (Figure 3.9B). Therefore, the
positive and negative weighting compounds effectively canceled out, which resulted in S.
guaranitica being positioned around zero on PC2.
S. nemorosa was positioned negatively on PC1. Some compounds from S. nemorosa (i.e.
β-caryophyllene, hentriacontane, 11-n-decyltetracosane, tetratriacontane, and hextriacontane)
were weighted positively, while other compounds (i.e. β-cubebene, an alkene, vitamin E, a
phenyl, and sitosterol) were weighted negatively on PC1. To explain the overall negative
positioning of this species on PC1, the mean-centered data were considered (Figure 3.12).
Due to the process of mean centering, compounds from other Salvia species were
introduced. For the replicate of S. nemorosa in Figure 3.12, compounds from S. officinalis (e.g.,
thujone, camphor, and epimanool) had a negative contribution in the mean-centered data. When
multiplied by the positive weighting compounds at these retention times in the eigenvector for
PC1, the product was negative. Since there were more negative loading compounds than
positive, S. nemorosa was positioned negatively on PC1 in the scores plot.
S. nemorosa was positioned around zero on PC2. β-caryophyllene was weighted
negatively while the other compounds present in S. nemorosa were weighted positively on PC2.
The positioning of S. nemorosa could be explained again with reference to the mean-centered
data. The abundance of all compounds was minimal in comparison to the abundance of the
internal standard, which, as previously mentioned, had both positive and negative contributions
in all but one replicate of S. nemorosa. With the exception of this replicate, the internal standard
was loaded both positively and negatively when multiplied by the eigenvector for PC2. In

76
 

1.00E+06

Abundance

A

0.00E+00

35 37

-1.00E+06

45

Retention Time (minutes)

3
1.20E+05

31

Loadings

B

0.00E+00

45
35
-1.20E+05
3

37
Retention Time (minutes)

31

Figure 3.12: (A) Mean-centered total ion chromatogram of one replicate of S. nemorosa and (B)
loadings for the same replicate of S. nemorosa on principal component 1 based on chemical
profiles obtained by gas chromatography-mass spectrometry with 35=thujone, 37=camphor, and
45=epimanool.
77
 

general, the positive and negative loadings canceled out, which resulted in S. nemorosa being
positioned around zero on PC2.
The replicate of S. nemorosa discussed in Figure 3.8, labeled replicate 1A, was
positioned more negatively than the others in the scores plot. After alignment was performed, the
internal standard in this replicate displayed drift in retention time. Therefore, this replicate had a
lower than average abundance of the internal standard at 24.46 min such that, on mean centering,
there was a negative contribution at this retention time. When multiplied by the eigenvector for
PC2, this peak was loaded negatively and had a greater contribution in this replicate than in the
other replicates of S. nemorosa. As a result, this replicate was positioned most negatively among
the other replicates of S. nemorosa on PC2 in the scores plot. Therefore, the spread observed in
PC2 for S. nemorosa was due to the misalignment of the internal standard.
S. splendens was positioned negatively on PC1. Several compounds from S. splendens
(i.e. nonacosane, hentriacontane, tetratriacontane, hexatriacontane, and oleanolic acid) were
weighted positively, while other compounds from S. splendens (i.e. an alkene, squalene, vitamin
E, 8-epi-salviarin, a cardenolide, sitosterol, and α-amyrin) were weighted negatively on PC1
(Figure 3.9A). To better explain the negative positioning of S. splendens on PC1, the meancentered data were investigated. As previously described, compounds from other species (i.e.
thujone, camphor, and epimanool from S. officinalis) were introduced during the process of mean
centering. In the mean-centered data for S. splendens, these compounds had a negative
contribution. When multiplied by the eigenvector for PC1, these compounds were loaded
negatively. Overall, the negatively loading compounds contributed more than those loading
positively such that S. splendens was positioned negatively on PC1. S. splendens was positioned

78
 

positively on PC2 in the scores plot since all compounds from this species were weighted
positively in the loadings plot for PC2 (Figure 3.9B).
To statistically assess the positioning of S. divinorum compared to the other four Salvia
species in the scores plot, two-tailed Student’s t-tests were performed on the mean scores of all
species for both PC1 and PC2, at the 99% confidence level. The unequal variance Student’s ttest was used. The null hypothesis, H0, tested was that the mean scores of the two species tested
were equal, while the alternative hypothesis, H1, stated that the mean scores of the two species
were statistically different. Table 3.8 shows the absolute value of the calculated t values and the
critical t values for comparisons of S. divinorum and the other four Salvia species on both PC1
and PC2 [12]. For all comparisons, the absolute value of tcalculated was greater than tcritical.
Therefore, S. divinorum was statistically differentiated from the remaining four species on both
PC1 and PC2, at the 99% confidence level.
Due to the close proximity of S. guaranitica, S. nemorosa, and S. splendens in the scores
plot, Student’s t-tests were also performed on pair-wise comparisons of these species to
determine if they could be statistically differentiated (Table 3.9). S. guaranitica was statistically
differentiated from S. splendens on both PC1 and PC2 at the 99% confidence level. S. nemorosa
and S. guaranitica were statistically differentiated on PC1 at the 99% confidence level. These
species were not statistically differentiated on PC2 at the 99% or the 95% confidence level, but
were differentiated at the 90% confidence level. S. nemorosa and S. splendens were statistically
differentiated on PC2 at the 99% confidence level; however, these species were not statistically
differentiated on PC1 at the 99%, 95%, and 90% confidence level. S. nemorosa and S. splendens
were positioned closely on the scores plot due to their chemical similarity. S. nemorosa and S.

79
 

Table 3.8: Absolute value of tcalculated and tcritical values calculated using two-tailed Student’s ttests performed on the mean scores of principal component 1 (PC1) and principal component 2
(PC2) for comparisons (n=18) of S. divinorum to the other Salvia species*
PC1
PC2
Comparisons of
Significant |t
Significant
tcritical
calculated|
S. divinorum to |tcalculated| tcritical
Difference
Difference
S. guaranitica 13.53
2.947
Yes
43.14
2.947
Yes
22.04
3.106
Yes
17.94
3.250
Yes
S. nemorosa
13.29
3.355
Yes
28.81
3.012
Yes
S. officinalis
20.28
3.106
Yes
58.80
3.055
Yes
S. splendens
*99% confidence interval [12]

80
 

Table 3.9: Absolute value of tcalculated and tcritical values calculated using two-tailed Student’s
t-tests performed on the mean scores of principal component 1 (PC1) and principal component 2
(PC2) for comparisons (n=18) between S. guaranitica, S. nemorosa, and S. splendens
PC1
PC2
Comparisons

|tcalculated|

tcritical

1.753*
S. nemorosa 1.728
to
S. splendens
3.106**
S. nemorosa 13.31
to
S. guaranitica
3.055**
S. guaranitica 11.40
to
S. splendens
*90% confidence interval [12]
**99% confidence interval [12]

Significant
Difference

|tcalculated|

tcritical

Significant
Difference

No

13.97

3.106**

Yes

Yes

2.272

1.860*

Yes

Yes

31.93

3.106**

Yes

81
 

splendens both contained an alkene, nonacosane, hentriacontane, vitamin E, tetratriacontane,
sitosterol, and hextriacontane.
Although Student’s t-tests demonstrated that S. divinorum could be differentiated from
four related Salvia species on PC1 and PC2 at the 99% confidence level, the scores plot only
accounted for 77% of the total variance. Therefore, PPMC coefficients, which assess similarity
between entire TICs based on a point-by-point comparison, were calculated between each Salvia
species, as well as between replicates of each species.
Mean PPMC coefficients calculated between S. divinorum and the remaining four Salvia
species were greater than 0.8, which indicated strong correlation (Table 3.10). However, the
range of PPMC coefficients between S. divinorum and the other four species indicated moderate
to strong correlation, except for the comparison of S. divinorum to S. guaranitica and to S.
splendens, which indicated strong correlation. These correlations were due to chemical
similarities between S. divinorum and the other four species. For example, all five species
contained hentriacontane, tetratriacontane, and sitosterol. S. divinorum, S. guaranitica, and S.
splendens all contained squalene. In addition, S. divinorum and S. splendens both contained
nonacosane. Furthermore, the internal standard added to all five species increased the correlation
between the species. This issue needs to be further investigated in the future.
S. divinorum and S. nemorosa contained some common compounds (i.e. hentriacontane,
tetratriacontane, sitosterol, and nonacosane). However, the moderate correlation between these
species was due to a misaligned internal standard peak previously mentioned.
S. divinorum and S. officinalis also contained common compounds (i.e. hentriacontane,
tetratriacontane, sitosterol, nonacosane, squalene, and heptacosane). S. officinalis contained

82
 

Table 3.10: Mean, standard deviation, and range of Pearson product moment correlation
(PPMC) coefficients for comparisons (n=81) between Salvia species
Comparisons
Mean PPMC coefficient ±
Range of PPMC coefficient
standard deviation
S. divinorum to S. guaranitica
0.9537 ± 0.0188
0.9077 –0.9821
S. divinorum to S. nemorosa
0.9490 ± 0.0460
0.7295 –0.9856
S. divinorum to S. officinalis
0.8100 ± 0.0508
0.7382 –0.8938
S. divinorum to S. splendens
0.9477 ± 0.0157
0.8999 –0.9759
S. guaranitica to S. nemorosa
0.9491 ± 0.0487
0.7283–0.9845
S. guaranitica to S.officinalis
0.8128 ± 0.0516
0.7402–0.8964
S. guaranitica to S. splendens
0.9529 ± 0.0166
0.8984–0.9777
S. nemorosa to S. officinalis
0.8255 ± 0.0621
0.6624–0.9206
S. nemorosa to S. splendens
0.9500 ± 0.0420
0.7925–0.9848
S. officinalis to S. splendens
0.8202 ± 0.0521
0.7454–0.9072

83
 

many early-eluting compounds that were not present in S. divinorum, which resulted in the
moderate correlation.
In addition, mean PPMC coefficients and ranges of PPMC coefficients were calculated
between pair-wise comparisons of replicates of S. guaranitica, S. nemorosa, and S. splendens
since these species were closely positioned on the scores plot. All mean PPMC coefficients were
greater than 0.9, which indicated strong correlation that was due to chemical similarities among
these species. However, the range of PPMC coefficients indicated a moderate to strong
correlation for S. nemorosa to S. guaranitica and to S. splendens, which was again due to the
misalignment of the internal standard peak in S. nemorosa.
S. divinorum was visually discriminated from the four Salvia species based on TICs.
However, PPMC coefficients showed that S. divinorum was moderately to strongly correlated to
the other four Salvia species; however, the internal standard added to all species artificially
increased the correlation coefficients. Using PCA and Student’s t-tests, S. divinorum was
discriminated from the other four species. This apparent contradiction can also be explained by
considering how PPMC coefficients and the PCA scores plot were calculated. Both PPMC
coefficients and the PCA scores plot were calculated using entire TICs; however, PPMC
coefficients were calculated between two TICs at a time based on a point-by-point comparison of
all data points. PCA magnified the differences between species and was used to identify the
most variable compounds among the species. Therefore, the PCA scores plot only accounted for
a certain amount of variance among the samples and as little similarity as possible. In addition,
PPMC coefficients were insensitive to relative abundance differences, while PCA was sensitive
to abundance differences. Therefore, the positioning of the species on the scores plot was
affected by relative abundance differences; however, the calculation of the PPMC coefficients

84
 

was not. Overall, this research showed that PCA was helpful in identifying the discriminating
factors among the entire data set and PCA along with Student’s t-tests discriminated S.
divinorum from the other four species despite all the similarities among the TICs of the species,
which was shown with PPMC coefficients.

3.4 Summary
S. divinorum and four related Salvia species were extracted by rotary agitation for 16
hours with dichloromethane and analyzed by GC-MS. Following analysis, TICs were
background subtracted, truncated, smoothed, retention time aligned, and normalized. PPMC
coefficients were calculated for pair-wise comparisons of replicate injections of each extract and
for all replicates of each species. In general, moderate to strong correlation among injection and
extraction replicates indicated acceptable precision in the extraction, instrumental analysis, and
data pretreatment procedures.
Using TICs, S. divinorum was visually discriminated from the four Salvia species. Using
PCA, each Salvia species was visually discriminated from S. divinorum on the scores plot.
Student’s t-tests, calculated on the mean scores on both PCs, showed S. divinorum was
statistically differentiated from the four other Salvia species at the 99% confidence level. PPMC
coefficients were also calculated between S. divinorum and the other Salvia species. Moderate to
strong correlations were observed, which indicated chemical similarity between S. divinorum and
the other four Salvia species. However, the internal standard added to each species increased the
correlation between species. Therefore, S. divinorum was discriminate from the other Salvia
species using TICs, PCA, and Student’s t-tests.

85
 

References

86
 

References
[1] Jermain JD, Evans HK. Analyzing Salvia divinorum and its active ingredient salvinorin A
utilizing thin layer chromatography and gas chromatography/mass spectrometry. J Forensic Sci
2009; 54:1–5.
[2] Bodnar Willard MA. Forensic analysis of Salvia divinorum and related Salvia species using
chemometric procedures. MS Thesis (2010).
[3] Radulescu V, Chiliment S, Oprea E. Capillary gas chromatography-mass spectrometry of
volatile and semi-volatile compounds of Salvia officinalis. Journal of Chromatography A 2004;
1027:121–6.
[4] Ulubelen A, Topcu G, Sönmez U, Eriş C. Terpenoids from Salvia nemorosa. Phytochemistry
1994; 35:1065–7.
[5] Carrer RP, Vanderlinde R, Dutra S, Marcon A, Echeverrigaray S. Essential oil variation
among Brazilian accessions of Salvia guaranitica L. Flavour Fragr. J. 2007; 22:430–4.
[6] Vallverdú C, Vila R, Lorenzo D, Paz D, Dellacassa E, Davies P, Villamil J, Tomi F,
Casanova J, Cañigueral S. Composition of the essential oil of cultivated Salvia guaranitica from
Uruguay 2005; 20:421–4.
[7] Fontana G, Savona G, Rodríguez B, Dersch C, Rothman R, Prisinzano T. Synthetic studies of
neoclerodane diterpenoids from Salvia splendens and evaluation of opioid receptor affinity.
Tetrahedron 2008; 64:10041–8.
[8] Johnson KJ, Wright BW, Jarman KH, Synovec RE. High-speed peak matching algorithm for
retention time alignment of gas chromatographic data for chemometric analysis. Journal of
Chromatography A 2003; 996:141–55.
[9] Walker JB, Sytsma KJ, Treutlein J, Wink M. Salvia (Laminaceae) is not monophyletic:
implications for the systematics, radiation, and ecological specializations of Salvia and tribe
Mentheae. American Journal of Botany 2004; 91:1115–25.
[10] Tychonievich, J. Salvia clades. Personal Communication. 2009.
[11] Medana C, Massolino C, Pazzi M, Baiocchi C. Determination of salvinorins and
divinatorins in Salvia divinorum leaves by liquid chromatography/multistage mass spectrometry.
Rapid Commun Mass Sp 2006; 20:131–6.
[12] Adam C. Essential mathematics and statistics for forensic science. Hobken, NJ: WileyBlackwell 2010.

87
 

Chapter 4: Association and Discrimination of Salvia divinorum from Related Salvia
Species Based on Chemical Fingerprints Generated Using Liquid ChromatographyMass Spectrometry
4.1 Introduction
Gas chromatography-mass spectrometry (GC-MS) is limited to the analysis of volatile
compounds that are thermally stable within normal operating temperatures of the instrument.
A chemical fingerprint of Salvia divinorum generated using GC-MS would then exclude
several non-volatile compounds. Hence, in this research, liquid chromatography-mass
spectrometry (LC-MS) was used to analyze the non-volatile compounds in S. divinorum and
related Salvia species. Chemical fingerprints generated using LC-MS are expected to be
complimentary to those generated using GC-MS. However, LC-MS may potentially offer
more discriminating information because some plant species are known to contain several
non-volatile compounds [1].
This chapter describes the discrimination of S. divinorum from the other Salvia
species based on their polar, non-volatile chemical compounds using LC-MS. Firstly,
extraction solvents and ionization modes were investigated using S. divinorum as the model
plant species. Then, S. divinorum, S. guaranitica, S. nemorosa, S. officinalis, and S. splendens
were extracted using a rotary agitator for 16 hours and analyzed using LC-MS and LC2

tandem mass spectrometry (MS ). Resulting total ion chromatograms (TICs) were truncated,
retention time aligned, and normalized. Mean Pearson product moment correlation (PPMC)
coefficients were then used to evaluate the precision of extraction and analysis procedures
and to evaluate the association of S. divinorum extracts and discrimination of S. divinorum
from the other four Salvia species. Principal components analysis (PCA) was used to evaluate
the discrimination of S. divinorum from the other four Salvia species in the scores plot.

88

Loadings plots were used to identify the compounds contributing most to the variance among
the species. Student’s t-tests were used to statistically evaluate the association and
discrimination in the PCA scores plot.

4.2 Material and Methods
4.2.1 Investigation of Extraction Solvent
Using leaves of S. divinorum, a preliminary study was conducted comparing two
extraction solvents: 100% acetonitrile (HPLC grade, EMD Chemical Inc., Darmstadt,
Germany) and a 3:3:2, by volume, mixture of acetonitrile (EMD Chemical Inc.), 2-propanol
(ACS grade, Sigma-Aldrich, St. Louis, MO), and water (HPLC grade, J. T. Baker,
Phillipsburg, NJ). Acetonitrile, or mixtures containing acetonitrile, have been used previously
for the extraction of S. divinorum and other plant materials [2-5]. In addition, acetonitrile is
more compatible with LC systems than other solvents used for GC analysis such as
dichloromethane.
All glassware used for the extraction was acid-washed following standard procedures.
Dried Salvia divinorum leaves were purchased from a local supplier (Frivolity Kingdom,
Jackson, MI). Leaves were stored in their original packaging at room temperature in complete
darkness until needed for analysis. A rapid and simple extraction was performed since these data
were only qualitatively analyzed to investigate extraction solvent. Approximately 0.2 g dried S.
divinorum leaves were placed in a beaker and 10 mL acetonitrile were added. After 5 minutes,
the extract was filtered using the experimental set-up described in Chapter 3 (Figure 3.1). The
beaker containing the extract was then rinsed with 5 mL acetonitrile and filtered. The extract and
rinses were collected, and then evaporated to dryness using nitrogen gas (Airgas, Independence,

89

OH) and a water circulating bath (NESLAB endocal RTE-9B, Newington, NH) at 35 °C. The
resulting extract was then stored at 2 °C until analysis. This procedure was repeated with new S.
divinorum leaves using the acetonitrile/2-propanol/water mixture as the extraction solvent.

4.2.2 Investigation of Liquid Chromatography-Mass Spectrometry Ionization Mode
Prior to LC-MS analysis, extracts were reconstituted in 0.58 mL of the appropriate
solvent (acetonitrile or acetonitrile/2-propanol/water). To assess injection precision, progesterone
(Sigma Chemical Company, St. Louis, MO) was added as an internal standard (0.00119 M) only
to the sample extracted using acetonitrile. However, due to the fixed volume injection loop of the
instrument, the investigation of injection precision by this method was later deemed unnecessary;
therefore, the addition of the internal standard to subsequent extracts was discontinued. The two
extracts were analyzed using an Applied Biosystems Sciex 3200 Q TRAP LC-MS system (Foster
City, CA) equipped with Analyst Software (version 1.4.2) and introduced via a Shimadzu
SILHTC autosampler (Columbia, MD), maintained at 15 °C, with a 5 µL fixed volume injection
loop. An Ascentis Express C-18 column (5.0 cm x 2.1 mm x 2.7 µm, Sigma-Aldrich, St Louis,
MO) was used and was maintained at 50 °C during analysis. The mobile phase was composed of
10 mM ammonium acetate (ACS reagent, J. T. Baker, Phillipsburg, NJ) in water (Milli-Q,
Billerica, MA), and methanol (HPLC grade, Sigma-Aldrich, St. Louis, MO), with a constant
flow rate of 0.4 mL/min. The methanol gradient was 1% at 1 minute, increasing linearly to 99%
until 20 minutes. From 20–27 minutes, the methanol gradient was held at 99%. Immediately
after, the methanol gradient was reduced back to 1% and held for approximately 3 minutes.
Following separation, compounds were ionized using electrospray ionization (ESI) with a
source temperature of 550 °C. Both positive and negative ionization modes were investigated.

90

The ion spray voltage was 5500 V for positive mode and -4500 V for negative mode. The
declustering potential was 45 V for positive mode and -45 V for negative mode. The mass
analyzer was a hybrid triple quadrupole/linear ion trap. Nitrogen was used as the collision gas,
set to an approximate pressure of 3.3 × 10-5 torr. A dynamic fill time was used to enhance the
accuracy of the mass spectra. An enhanced mass spectrum (EMS) scan was collected from m/z
80–1200 with a scan rate of 1000 amu/s.

4.2.3 Extraction of Salvia Species
Following investigation of the extraction solvent and ionization mode, S. divinorum and
the four Salvia species (S. guaranitica, S. nemorosa, S. officinalis, and S. splendens) were
extracted for analysis. Salvia leaves were obtained from the same sources and extracted using the
procedure described in Chapter 3, section 3.2.1, except using acetonitrile (EMD Chemical Inc.)
as the extraction solvent rather than dichloromethane. This extraction procedure was repeated in
triplicate for each Salvia species, using different leaf material for each extraction. All extracts
were stored at 2 °C until analysis.

4.2.4 Liquid Chromatography-Mass Spectrometry Analysis
All extracts were reconstituted in 0.58 mL acetonitrile (EMD Chemical Inc.), shaken and
vortexed, and each extract was divided into three autosampler vials with inserts to allow for
replicate injections. This was necessary to minimize the risk of evaporative loss due to
puncturing the septum of the vial during analysis of replicates. Instrument parameters were
similar to those described in section 4.2.2 with some exceptions. The autosampler was
maintained at 5 °C during analysis to prevent sample degradation. The ion spray voltage was

91

5500 V and the declustering potential was 30 V. An EMS scan was collected in positive mode
over the range m/z 80–1000 with a scan rate of 4000 amu/s.
2

The Salvia extracts were also analyzed by MS to aid in compound identification. An
information dependent acquisition (IDA) with an EMS scan and enhanced product ion (EPI) scan
was performed. The IDA was programmed to select the two most abundant fragments from the
EMS scan and perform the EPI scan. The parameters in section 4.2.2 were used with some
additions. Due to the presence of masses at m/z 87.000, 141.000, 158.000, and 214.1000 in the
background, these ions were excluded from tandem experiments. For the EMS scan, the collision
energy was set at 10 V with a collision energy spread of 0 V. For the EPI scans, the collision
energy was set at 45 V with a collision energy spread of 15 V.

4.2.5 Data Pretreatment
Each extract was analyzed in triplicate, which resulted in a total of nine TICs for each
Salvia species. All TICs were pretreated prior to data analysis. As a result of using a dynamic fill
time during mass analysis, TICs had different numbers of data points. This would cause
problems in subsequent PCA because the difference would be identified as the major source of
variance among the extracts. As a result, it was necessary to standardize the number of data
points across all TICs. To do this, TICs were first truncated to the smallest number of data points
(1511) among all Salvia species. An acetonitrile TIC with a similar number of data points (1512)
was then selected as the target chromatogram and truncated to 1511 data points. Using a
correlation optimized warping (COW) algorithm in LineUp™ (Version 3.0, InfoMetrix, Inc.,
Bothell, WA), truncated TICs were first aligned to the acetonitrile target TIC with a warp of 6
data points and segment size of 250 data points. At this stage, all TICs had the same number of

92

data points and these data were used in subsequent pretreatment procedures.
For retention time alignment, an average target chromatogram was created in Microsoft
Excel (Versions 2007 and 2010, Microsoft Corporation, Redmond, WA). Chapter 3, section
3.2.3, explains how the average target chromatogram was created. Total ion chromatograms were
aligned to the target chromatogram using the COW algorithm with a warp of 6 data points and
segment size of 65 data points. Following alignment, TICs were normalized using total area
normalization by species.

4.2.6 Data Analysis
Mean PPMC coefficients were calculated in Matlab (Version 7.11.0.584 R2010b, The
Mathworks, Natick, MA) on pretreated TICs for all pair-wise combinations of replicate
injections of the same extract and for replicates of each species to assess the precision of the
extraction and analysis procedures. Principal components analysis was then performed on the
aligned and normalized data using Matlab. Scores and loadings plots were generated in
Microsoft Excel. To statistically assess discrimination of Salvia divinorum from the four other
Salvia species, Student’s t-tests were performed in Microsoft Excel on the mean scores of each
species on both principal component 1 (PC1) and principal component 2 (PC2) at the 99%
confidence level. To further assess association and discrimination of S. divinorum from the four
Salvia species, PPMC coefficients were calculated on TICs for all pair-wise comparisons of
species.

93

4.3 Results and Discussion
4.3.1 Investigation of Extraction Solvent and Ionization Mode
Total ion chromatograms of S. divinorum extracted using acetonitrile and acetonitrile/2propanol/water, analyzed using both positive and negative ionization were visually compared.
Regardless of the extraction solvent used, S. divinorum analyzed in positive mode contained
compounds at abundances approximately an order of magnitude greater than those analyzed in
negative mode. Additionally, more compounds (i.e. salvinorins A–D and F) were identified in
positive mode than in negative mode [2, 3, 6]. Therefore, for all subsequent analyses, positive
ionization mode was used.
To determine the best extraction solvent, TICs of S. divinorum analyzed in positive mode
following an extraction using acetonitrile and acetonitrile/2-propanol/water were compared
(Figure 4.1). Salvinorins A, B, C, D, F, as well as five unidentified compounds (A–E), were
extracted at relatively similar abundances by both solvents [2, 3, 6]. Since no obvious differences
in compounds extracted were apparent between the two solvents, practicality of the extraction
procedure was considered. In a forensic laboratory, efficient analyses are necessary to increase
sample throughput. Due to its polarity, water is difficult to evaporate and therefore the
acetonitrile/2-propanol/water was not considered further. For all subsequent extractions,
acetonitrile was used as the extraction solvent and all extracts were analyzed in positive mode.

4.3.2 Provisional Identification of Compounds Present in Each Species
Representative TICs of the five Salvia species extracted in acetonitrile and analyzed in
positive mode are shown in Figures 4.2–4.6. Provisional identifications for compounds present

94

2.50E+09

A
6
8

Abundance

11

7

10

4, 5
3
2
9
1
0.00E+00
0

Retention Time (minutes)

2.50E+09
11

Abundance

6

30

B

4,5
3
2 7

10
9

1
0.00E+00
0

Retention Time (minutes)

30

Figure 4.1: Total ion chromatograms of S. divinorum extracted in (A) acetonitrile with
progesterone added as an internal standard (IS) and (B) acetonitrile/2-propanol/water. Both
extracts were analyzed using positive ionization mode. Provisional identifications, based on
literature sources, are as follows: 1=salvinorin B, 2=salvinorin D, 3=salvinorin F, 4=unidentified
compound A, 5=unidentified compound B, 6=salvinorin A, 7=salvinorin C, 8=progesterone (IS),
9=unidentified compound C, 10=unidentified compound D, and 11=unidentified compound E [2,
3, 6, 7].
95

1.40E+09
9

Abundance

6

8

1
2,3
4
0.00E+00
0

5
7

Retention Time (minutes)

30

Figure 4.2: Total ion chromatogram of Salvia divinorum extracted in acetonitrile and analyzed in
positive mode using liquid chromatography-mass spectrometry. Provisional identifications based
on literature sources are as follows: 1=salvinorin B, 2=terpenoid (molecular mass= m/z 406),
3=terpenoid (molecular mass= m/z 378), 4=salvinorin D, 5=salvinorin F, 6=salvinorin A,
7=salvinorin C, 8=sesquiterpenoid, and 9=unidentified compound (molecular mass= m/z 426) [2,
3, 6]. See Appendix B, Table B.1, for more detailed information.

96

Abundance

1.40E+09

11

10
12 13
0.00E+00
0

14

8

Retention Time (minutes)

15 16

30

Figure 4.3: Total ion chromatogram of Salvia guaranitica extracted in acetonitrile and analyzed
in positive mode using liquid chromatography-mass spectrometry. Provisional identifications
based on literature sources are as follows: 8=sesquiterpenoid, 10=diterpenoid (molecular
mass=m/z 340), 11=terpenoid (molecular mass=m/z 398), 12= terpenoid (molecular masses=m/z
340, 398), 13=terpenoid (molecular mass=m/z 398), 14=oleanolic/ursolic acid, 15=vitamin E,
and 16=unidentified compound [8–12]. See Appendix B, Table B.2, for more detailed
information.

97

Abundance

1.40E+09

8
14

0.00E+00
0

15
17

Retention Time (minutes)

30

Figure 4.4: Total ion chromatogram of Salvia nemorosa extracted in acetonitrile and analyzed in
positive mode using liquid chromatography-mass spectrometry. Provisional identifications based
on literature sources are as follows: 8=sesquiterpenoid, 14=oleanolic/ursolic acid, 15=vitamin E,
and 17=unidentified compound [8–12]. See Appendix B, Table B.3, for more detailed
information.

98

1.40E+09
19

Abundance

20

8
15
14 21
18

0.00E+00
0

Retention Time (minutes)

30

Figure 4.5: Total ion chromatogram of Salvia officinalis extracted in acetonitrile and analyzed in
positive mode using liquid chromatography-mass spectrometry. Provisional identifications based
on literature sources are as follows: 8=sesquiterpenoid, 14=oleanolic/ursolic acid, 15=vitamin E,
18=3-O-methylkaempferol, 19=carnosol and carnosic acid, 20=methyl carnosate/rosmanol/
epirosmanol, and 21=unidentified compound [8–12, 13–17]. See Appendix B, Table B.4, for
more detailed information.

99

Abundance

1.40E+09

15
9
22-24
25
0.00E+00
0

8
14

Retention Time (minutes)

30

Figure 4.6: Total ion chromatogram of Salvia splendens extracted in acetonitrile and analyzed in
positive mode using liquid chromatography-mass spectrometry. Provisional identifications based
on literature sources are as follows: 8=sesquiterpenoid, 9=unidentified compound (molecular
mass=m/z 426),14=oleanolic/ursolic acid, 15=vitamin E, 22=diterpenoid (molecular mass=m/z
400), 23=diterpenoid (molecular mass=m/z 342), 24=diterpenoid (molecular mass=m/z 362), and
25=phenolic methyl ether and an unidentified compound (molecular masses=m/z 406, 548) [8–
12]. See Appendix B, Table B.5, for more detailed information.

100

in each species with molecular masses and major fragmentation ions are listed in Appendix B.
Some peaks are identified as more than one compound, due to compounds being isomers (e.g.,
oleanolic and ursolic acid, as well as rosmanol and epirosmanol [15]), compounds having the
same molecular ion and major fragment ions (e.g., rosmanol, epirosmanol, and methyl carsonate
[17]), or compounds being poorly resolved (e.g., compounds 19 and 25).
S. nemorosa and S. officinialis belong to Salvia clade I, while S. divinorum, S.
guaranitica, and S. splendens belong to Salvia clade II [18–20]. Species in different clades
contained common compounds; therefore, classification of the species into clades based on the
non-volatile compounds was also difficult. For example, all five Salvia species contained a
sesquiterpenoid with a molecular mass of m/z 264. S. guaranitica, S. nemorosa, S. officinalis,
and S. splendens all contained oleanolic and/or its isomer, ursolic acid and vitamin E [8–12]. In
addition, compounds were common among some species in the same clade but not all species in
that clade. For example, S. divinorum and S. splendens, which are both in Salvia clade II,
contained an unidentified compound with a molecular mass of m/z 426; however, S. guaranitica,
which also belongs to Salvia clade II, did not contain this compound [18]. Furthermore, it is
noteworthy that S. divinorum was the only species to contain salvinorins A, B, C, D, and F and
TICs of S. divinorum were visually discriminated from the other four Salvia species [2, 3, 6].

4.3.3 Investigation of Extraction and Analysis Precision Using Pearson Product Moment
Correlation Coefficients
The mean, standard deviation, and range of PPMC coefficients were calculated for
replicate injections (n=3) of each extract to demonstrate precision in the instrumental analysis
procedure (Table 4.3). However, since data pretreatment was performed prior to calculating

101

Table 4.3: Mean, standard deviation, and range of Pearson product moment correlation (PPMC) coefficients for replicate injections of
the same extract
Mean PPMC coefficients ± standard deviation
Range of PPMC coefficients
Species

Extract 1 (n=3)

Extract 2 (n=3)

Extract 3 (n=3)

Extract 1 (n=3)

Extract 2 (n=3)

Extract 3 (n=3)

S. divinorum

0.9577 ± 0.0192

0.9801 ± 0.0049

0.9563 ± 0.0007

0.9537–0.9786

0.9772–0.9858

0.9558–0.9571

S. guaranitica

0.9641 ± 0.0053

0.9379 ± 0.0254

0.9530 ± 0.0147

0.9585–0.9691

0.9171–0.9663

0.9423–0.9698

S. nemorosa

0.7601 ± 0.0321

0.9496 ± 0.0229

0.8281 ± 0.1086

0.7230–0.7794

0.9257–0.9714

0.7647–0.9535

S. officinalis

0.9718 ± 0.0014

0.9785 ± 0.0046

0.9676 ± 0.0097

0.9705–0.9733

0.9747–0.9836

0.9564–0.9737

S. splendens

0.8679 ± 0.0350

0.9589 ± 0.0153

0.9014 ± 0.0393

0.8293–0.8976

0.9424–0.9727

0.8761–0.9467

102

PPMC coefficients, precision of these procedures was also reflected in the coefficients.
All mean PPMC coefficients for injection replicates were 0.9 or greater, with the
exception of Extracts 1 and 3 of S. nemorosa and Extract 1 of S. splendens. The ranges of
coefficients were all greater than 0.8 except Extract 1 and 3 of S. nemorosa. Ideally replicates
are expected to have a PPMC coefficient of 1.00. However, the retention time alignment of S.
nemorosa and S. splendens was compromised to improve the alignment of the other species.
This resulted in retention time misalignments of the sesquiterpenoid and oleanolic and/or ursolic
acid in S. nemorosa and S. splendens, which lowered the mean PPMC coefficients among
replicates of these species. All standard deviations were low (≤ 0.04) except for Extract 3 of S.
nemorosa, for which the standard deviation was 0.1086. This higher standard deviation was due
to misalignments of the sesquiterpenoid and oleanolic and/or ursolic acid as previously
discussed. Thus, overall, there was acceptable precision in the instrumental analysis and data
pretreatment procedures.
To demonstrate precision in the extraction procedure, in addition to the instrumental
analysis and data pretreatment procedures, the mean, standard deviation, and range of PPMC
coefficients were also calculated for all extracts of each species (n=36). All mean PPMC
coefficients were greater than 0.8, which indicated strong correlation between extracts of the
same species (Table 4.4). The ranges of the coefficients were greater than 0.8 with the exception
of S. nemorosa, which also had a higher standard deviation (0.0874) than the replicates of the
other species due to the retention time misalignments discussed previously. Overall, and despite
these misalignments, PPMC coefficients indicated acceptable precision in the extraction,
instrumental analysis, and data pretreatment procedures.

103

Table 4.4: Mean, standard deviation, and range of Pearson product moment correlation (PPMC)
coefficients for all extracts for each species
Species
Mean PPMC coefficient ± standard
Range of PPMC coefficients
deviation (n=36)
(n=36)
S. divinorum

0.9498 ± 0.0165

0.9210–0.9858

S. guaranitica

0.9402 ± 0.0199

0.9017–0.9698

S. nemorosa

0.8300 ± 0.0874

0.7214–0.9714

S. officinalis

0.9441 ± 0.0282

0.8853–0.9836

S. splendens

0.9041 ± 0.0377

0.8293–0.9727

104

4.3.4 Association and Discrimination of Salvia divinorum from Other Salvia Species
The scores plot (Figure 4.7) of PC1 and PC2 accounted for 82% of the total variance
among the Salvia species. S. officinalis was the only species positioned positively on PC1, while
S. divinorum was the only species positioned positively on PC2. S. guaranitica, S. nemorosa,
and S. splendens were positioned closely, negatively on both PC1 and PC2. Replicates of each
species were generally closely clustered in the scores plot on both PC1 and PC2, with the
exception of S. officinalis for which spread among replicates was observed. The positioning of
all species, and spread among replicates of S. officinalis, in the scores plot could be explained
based on the loadings plots for PC1 and PC2 (Figure 4.8).
S. officinalis was positioned positively on PC1 because it contained all of the compounds
that were weighted positively on this PC (e.g. phenolic diterpenes, sesquiterpenoid, oleanolic
and/or ursolic acid, vitamin E, and an unidentified compound). While one compound present in
S. officinalis (3-O-methylkaempferol), was weighted negatively on PC1, its contribution was
minimal in comparison to the positively weighted compounds [13]. As a result, S. officinalis was
positioned positively on PC1 in the scores plot.
Compounds present in S. officinalis were weighted both positively (e.g., a
sesquiterpenoid, vitamin E, and an unidentified compound) and negatively (e.g., 3-Omethylkaempferol, phenolic diterpenes, and oleanolic and/or ursolic acid) on PC2. However,
since the contribution of the negatively weighted compounds was greater than that of the
positively weighted compounds, S. officinalis was positioned negatively on PC2.
The spread observed among replicates of S. officinalis was the result of variability in the
abundance of carnosol and carnosic acid, and methyl carnosate/rosmanol/epirosmanol in

105

PC2 (22%)

4.00E+09

-4.00E+09

0.00E+00

4.00E+09

-4.00E+09
PC1 (60%)
S. divinorum
S. officinalis

S. guaranitica
S. splendens

S. nemorosa

Figure 4.7: Scores plot of principal component 1 versus principal component 2 showing
discrimination of S. divinorum from the other four Salvia species based on chemical profiles
obtained by liquid chromatography-mass spectrometry.

106

0.35
Loadings on PC1

A
20
19

8
14

0
10

9, 15–17, 21
15 15
15

6, 13, 25
3, 12, 24
2, 11, 18, 23

-0.35
0

Retention Time (minutes)

30

Figure 4.8: Loadings plots for (A) principal component 1 (PC1) and (B) principal component 2
(PC2) based on principal components analysis of the five Salvia species analyzed by liquid
chromatography-mass spectrometry with 1=salvinorin B, 2=terpenoid , 3=terpenoid,
4=salvinorin D, 5=salvinorin F, 6=salvinorin A, 7=salvinorin C, 8=sesquiterpenoid,
9=unidentified compound, 10=diterpenoid, 11=terpenoid, 12=terpenoid, 13=terpenoid,
14=oleanolic/ursolic acid, 15=vitamin E, 16=unidentified compound, 17=unidentified
compound, 18=3-O-methylkaempferol, 19=carnosol and carnosic acid, 20=methyl carnosate/
rosmanol/epirosmanol, 21=unidentified compound, 22=diterpenoid, 23=diterpenoid,
24=diterpenoid, 25=phenolic methyl ether and unidentified compound.

107

S. officinalis. The amount of carnosic acid and carnosol are known to change as leaves age and
therefore, may be the cause of the variability in abundance of these compounds [14–17, 21]. In
addition, carnosic acid can decompose into rosmanol, epirosomanol, and/or carnosol in the
presence of methanol [22]. As previously stated, methanol was used in the mobile phase to
perform the LC analysis. Since these compounds were weighted positively on the loadings plot
for PC1 and negatively on the loadings plot for PC2, variability in abundance of these
compounds affected the positioning of S. officinalis replicates. For instance, replicates with a
greater abundance of these compounds were positioned more positively on PC1 and more
negatively on PC2 than the other replicates.
S. divinorum was positioned slightly negatively on PC1. This species contained
salvinorin A and two terpenoids that were weighted negatively on PC1, as well as a
sesquiterpenoid and an unidentified compound that were weighted positively. The overall
negative positioning of S. divinorum on PC1 in the scores plot was a result of mean centering the
TICs, which was necessary for PCA.
To mean center a data set, the first step was to calculate the average abundance of all
sample TICs at each retention time. This average was then subtracted from the abundance at
each retention time in each TIC to generate the mean-centered data. Figure 4.9 shows the meancentered data for one replicate of S. divinorum.
Due to the process of mean centering, compounds from the other Salvia species were
introduced. For the replicate of S. divinorum in Figure 4.9A, phenolic diterpenes (compounds 19
and 20) from S. officinalis had a negative contribution in the mean-centered data. The meancentered data were then multiplied by the eigenvector for PC1, which generated the loadings,
and every data point was summed to generate the score for the sample on PC1. The eigenvector

108

8.00E+08

Abundance

A

0.00E+00
19 20

-8.00E+08
0

Retention Time (minutes)

2.50E+07

30

Loadings

B

0.00E+00

19
-2.50E+07
0

20
Retention Time (minutes)

30

Figure 4.9: (A) Mean-centered total ion chromatogram of S. divinorum and (B) loadings for S.
divinorum on principal component 1 based on chemical profiles obtained by liquid
chromatography-mass spectrometry with 19=carnosol and carnosic acid, and 20=methyl
carnosate/rosmanol/epirosmanol.
109

was positive at the retention times corresponding to the phenolic diterpenes. Therefore, when
multiplied by the mean-centered data, these compounds were loaded negatively (Figure 4.9B).
Overall, the negatively loading compounds contributed more than those loading positively such
that S. divinorum was positioned slightly negatively on PC1.
S. divinorum was positioned positively on PC2 in the scores plot mainly due to
salvinorins A, B, C, D, and F, a sesquiterpenoid, and an unidentified compound that were
weighted positively (Figure 4.8B). Two terpenoids from S. divinorum were weighted negatively
on PC2. However, since many more compounds in this species were weighted positively than
negatively, S. divinorum was positioned positively on PC2.
S. nemorosa was positioned negatively on PC1 in the scores plot. However, no
compounds from S. nemorosa were weighted negatively on the PC1 loadings plot. This apparent
contradiction in the positioning of S. nemorosa could be explained by investigating the meancentered data. Through the act of mean centering, the phenolic diterpenes from S. officinalis
were introduced and had a negative contribution. When multiplied by the eigenvector for PC1,
these compounds loaded negatively. Overall, the compounds loading negatively contributed
more than those loading positively such that S. nemorosa was positioned negatively on PC1 in
the scores plot.
S. nemorosa was also positioned negatively on PC2 due to the process of mean
centering. In addition to the phenolic diterpenes from S. officinalis, compounds from other
species were introduced into the mean-centered data for S. nemorosa (Figure 4.10 A). For
example, salvinorin A from S. divinorum, a terpenoid from S. guaranitica, and a phenolic
methyl ether from S. splendens contributed negatively in the mean-centered data at
approximately 12.3 minutes. When multiplied by the eigenvector for PC2 (Figure 4.10 B), these

110

2.50E+08

Abundance

A

0.00E+00

6, 13, 25

19 20
9, 15-17, 21

-2.50E+08
0

Retention Time (minutes)

7.00E+07

30

Loadings

B

19 20
0.00E+00

6, 13, 25
9, 15–17, 21

-7.00E+07
0

Retention Time (minutes)

30

Figure 4.10: (A) Mean-centered total ion chromatogram of S. nemorosa and (B) loadings
for S. nemorosa on principal component 2 based on chemical profiles obtained by liquid
chromatography-mass spectrometry with 6=salvinorin A, 9=unidentified compound,
13=terpenoid, 15=vitamin E, 16=unidentified compound, 17=unidentified compound,
19=carnosol and carnosic acid, 20=methyl carnosate/rosmanol/epirosmanol, and
21=unidentified compound.
111

compounds loaded negatively while the phenolic diterpenes loaded positively. Overall, the
negative loading compounds contributed more than those loading positively such that S.
nemorosa was positioned negatively on PC2.
S. splendens was positioned negatively on PC1 in the scores plot also due to the process
of mean centering. In the mean-centered data, phenolic diterpenes from S. officinalis were
introduced and contributed negatively. When multiplied by the eigenvector for PC1, these
compounds loaded negatively, which resulted in the negative positioning of S. splendens on
PC1.
S. splendens was also positioned negatively on PC2 due to the process of mean centering
(Figure 4.11). In the mean-centered data for S. splendens, a phenolic methyl ether and
unidentified compound from S. splendens had a negative contribution at the same retention time
as salvinorin A and the unidentified compound from S. divinorum, respectively. The compounds
in the TIC of S. divinorum were more abundant than the compounds at the same retention time
in the TIC of S. splendens. As a result, the average abundance of all TICs at these retention
times was greater than the abundance at these retention times in S. splendens, which caused the
compounds at these retention times to have a negative contribution in the mean-centered data.
These compounds were weighted positively in the eigenvector for PC2 such that, when
multiplied by the mean-centered data, these compounds were loaded negatively. This resulted in
the negative positioning of S. splendens on PC2.
As observed in the scores plot, S. nemorosa and S. splendens were positioned closely on
both PC1 and PC2. This close positioning was due to chemical similarities between these two
species. Since S. nemorosa and S. splendens contained fewer and less abundant compounds than
the other three species, compounds from the other species were introduced into the

112

1.50E+08

Abundance

A

0.00E+00

9, 15–17, 21

6, 13, 25
-1.50E+08
0

Retention Time (minutes)

4.00E+07

30

Loadings

B

0.00E+00

6, 13, 25
-4.00E+07
0

9, 15–17, 21

Retention Time (minutes)

30

Figure 4.11: (A) Mean-centered total ion chromatogram of S. splendens, and (B) loadings for S.
splendens on principal component 2 based on chemical profiles obtained by liquid
chromatography-mass spectrometry with 6=salvinorin A, 9=unidentified compound,
13=terpenoid, 15=vitamin E, 16=unidentified compound, 17=unidentified compound,
21=unidentified compound, and 25=phenolic methyl ether and unidentified compound.

113

mean-centered data for S. nemorosa and S. splendens. Therefore, when the mean-centered data
for replicates of S. nemorosa and for replicates of S. splendens were multiplied by the
eigenvector for PC1, many of the same compounds were present. For example, both the loadings
for S. nemorosa and for S. splendens contained a terpenoid from S. guaranitica, salvinorin A
from S. divinorum, and phenolic diterpenes from S. officinalis. The close positioning of S.
nemorosa and S. splendens on PC2 could be explained in a similar manner.
In addition, there was minimal spread among replicates of S. nemorosa and S. splendens
in the scores plot; however, as previously stated, both these species had lower PPMC coefficients
due to misalignments of a sesquiterpenoid, and oleanolic and/or ursolic acid. This contradiction
between PPMC coefficients and PCA was reconciled when considering the differences between
these two procedures as previously discussed in Chapter 3, section 3.3.3. In addition, PPMC
coefficients were low because of the misalignments of oleanolic and/or ursolic acid and the
sesquiterpenoid. However, in general, these compounds did not contribute greatly to the
positioning of S. nemorosa and S. splendens on the scores plot.
S. guaranitica was positioned negatively on PC1 in the scores plot due to the process of
mean centering. In the mean-centered data for S. guaranitica, a terpenoid had a positive
contribution while phenolic diterpenes introduced into the mean-centered data from S.
officinalis, as well as a sesquiterpenoid, had a negative contribution. When the mean-centered
data were multiplied by the eigenvector for PC1, all these compounds loaded negatively, which
resulted in the negative positioning of S. guaranitica on PC1 in the scores plot.
S. guaranitica was also positioned negatively on PC2 in the scores plot. The positioning
of S. guaranitica could be explained using the mean-centered data. In the mean-centered data for
S. guaranitica, some compounds from S. guaranitica contributed positively while others

114

contributed negatively. In addition, two phenolic diterpenes from S. officinalis were introduced
into the mean-centered data and contributed negatively. When the mean-centered data were
multiplied by the PC2 eigenvector, the negative loading compounds contributed more than those
loading positively, which resulted in the negative positioning of S. guaranitica on PC2.
To statistically assess differentiation of S. divinorum from the other four Salvia species,
two-tailed Student’s t-tests were performed on the mean scores for both PC1 and PC2 at the 99%
confidence level. Variances between all comparisons were assumed to be unequal. The null
hypothesis, H0, tested was that the mean scores of any two species were equal while the
alternative hypothesis, H1, stated that the mean scores of any two species were statistically
different. Table 4.7 shows the absolute value of the calculated t values and critical t values for
comparison of S. divinorum and the other four Salvia species on PC1 and PC2 [23]. For all
comparisons, the absolute value of t calculated was greater than tcritical. Therefore, Student’s ttests showed S. divinorum was statistically differentiated from the remaining four species on both
PC1 and PC2, at the 99% confidence level.
Due to the close proximity of S. guaranitica, S. nemorosa, and S. splendens in the scores
plot, Student’s t-tests were also performed on each pair-wise comparison of these species to
determine if they could be statistically differentiated (Table 4.8) [23]. For all these comparisons,
the absolute value of tcalculated was greater than tcritical. Thus, despite close positioning in the
scores plot, S. guaranitica, S. nemorosa, and S. splendens were statistically differentiated from
each other on PC1 and PC2 at the 99% confidence level.

115

Table 4.7: Absolute value of tcalculated and tcritical values calculated using two-tailed Student’s ttests performed on the mean scores of principal component 1 (PC1) and principal component 2
(PC2) for comparisons (n=18) of S. divinorum to the other Salvia species*
PC1
PC2
Comparisons of
|tcalculated| tcritical Significant |tcalculated| tcritical Significant
S. divinorum to
Difference
Difference
41.29
2.947
Yes
38.47
3.106
Yes
S. guaranitica
11.97
2.947
Yes
34.83
3.106
Yes
S. nemorosa
32.89
3.355
Yes
25.47
2.921
Yes
S. officinalis
26.85
3.250
Yes
33.95
3.355
Yes
S. splendens
*99% confidence interval [23]

116

Table 4.8: Absolute value of tcalculated and tcritical values calculated from Student’s t-tests
performed on the mean scores of principal component 1 (PC1) and principal component 2 (PC2)
between S. guaranitica, S. nemorosa, and S. splendens (n=18)*
PC1
PC2
Comparisons
Significant |t
Significant
|tcalculated| tcritical
calculated| tcritical
Difference
Difference
7.380
3.250
Yes
7.210
3.106
Yes
S. nemorosa
to
S. splendens
23.72
2.977
Yes
5.266
2.947
Yes
S. nemorosa
to
S. guaranitica
28.07
3.169
Yes
14.61
3.106
Yes
S. guaranitica
to
S. splendens
*2-tailed t-test calculated at the 99% confidence interval [23]

117

Therefore, S. divinorum was statistically differentiated from the remaining four species
in the scores plot on both PC1 and PC2 at the 99% confidence level using both the GC-MS data
(Chapter 3, section 3.3.3) and LC-MS data. Using the LC-MS data, S. guaranitica, S. nemorosa,
and S. splendens were all statistically differentiated in the scores plot on both PC1 and PC2 at a
99% confidence level; however, this did not occur when extracts the same species were analyzed
by GC-MS. Therefore, the non-volatile compounds analyzed and detected by LC-MS offered
greater discrimination of the species than the more volatile compounds analyzed and detected by
GC-MS.
Although Student’s t-tests demonstrated that S. divinorum could be differentiated from
four related Salvia species on PC1 and PC2 at the 99% confidence level, the scores plot only
accounted for 82% of the total variance. Therefore, PPMC coefficients, which assess similarity
between pairs of samples based on a point-by-point comparison, were calculated between each
Salvia species.
Mean PPMC coefficients calculated for the comparison of S. divinorum to S. guaranitica
and to S. officinalis were less than 0.5, which indicated weak correlation (Table 4.9). For the
comparison of S. divinorum to S. nemorosa and to S. splendens, mean PPMC coefficients were
between 0.5 and 0.8, which indicated moderate correlation. These correlations were due to
chemical differences between S. divinorum and the other four species. For example, S.
divinorum was the only species to contain salvinorins A, B, C, D, and F. In addition, there was
only one compound, a sesquiterpenoid, which was common among all five Salvia species. S.
divinorum and S. splendens also both contained an unidentified compound, which resulted in the
higher PPMC coefficients between these species than between S. divinorum and the other three
species.

118

Table 4.9: Mean, standard deviation, and range of Pearson product moment correlation (PPMC)
coefficients for comparisons (n=81) between Salvia species
Comparisons
Mean PPMC coefficient ±
Range of PPMC coefficient
standard deviation
S. divinorum to S. guaranitica
0.4068 ± 0.0581
0.2788–0.5098
S. divinorum to S. nemorosa
0.5469 ± 0.0762
0.3750–0.6893
S. divinorum to S. officinalis
0.3762 ± 0.0721
0.2506–0.5177
S. divinorum to S. splendens
0.6432 ± 0.0602
0.5348–0.7561
S. guaranitica to S. nemorosa
0.4330 ± 0.0656
0.3119–0.5843
S. guaranitica to S.officinalis
0.2534 ± 0.0592
0.1417–0.3739
S. guaranitica to S. splendens
0.5937 ± 0.0601
0.4653–0.7119
S. nemorosa to S. officinalis
0.6126 ± 0.0880
0.4479–0.7783
S. nemorosa to S. splendens
0.7628 ± 0.0418
0.6816–0.8662
S. officinalis to S. splendens
0.5048 ± 0.0773
0.3733–0.6237

119

Mean PPMC coefficients calculated based on the LC-MS data were below 0.8, while
mean PPMC coefficients calculated based on the GC-MS data (Chapter 3, section 3.3.3) were
greater than 0.8. However, the internal standard, added to all five species analyzed using GCMS, increased the correlation between the species. Therefore, this issue needs to be further
investigated in the future.
In addition, the mean PPMC coefficient and range of PPMC coefficient were calculated
between S. nemorosa and S. splendens since these species were positioned closest to one another
on the scores plot. These species had a mean PPMC coefficient of 0.7628 ± 0.0418, which
indicated a moderate correlation. However, the range of PPMC coefficients (0.6816–0.8662)
between these species indicated a moderate to strong correlation due to chemical similarities.
These species both contained a sesquiterpenoid, oleanolic and/or ursolic acid, and vitamin E.
Also, these species contained fewer and less abundant additional compounds compared to the
other three Salvia species.
PPMC coefficients showed that S. splendens and S. nemorosa were moderate to strongly
correlated, which was supported by the close positioning of these species in the scores plot.
However, as previously stated, these two species were differentiated on both PCs using
Student’s t-tests. This apparent contradiction was reconciled when considering the differences
between these procedures. PPMC coefficients were calculated between two TICs based on a
point-by-point comparison of all data points. Likewise, PCA was calculated using entire TICs.
However, Student’s t-tests were calculated on the mean scores of each species on PC1 and PC2;
therefore, essentially minimizing the comparison between the two species to a one point
comparison on both PCs. Therefore, this research showed that despite similarities among these

120

species, which was shown with PPMC coefficients and PCA scores plot, Student’s t-tests were
helpful in discriminating these two species.
In conclusion, TICs of S. divinorum were visually discriminated from the other four
Salvia species; however, more salvinorin compounds were identified in the extracts analyzed
using LC-MS than GC-MS. PPMC coefficients showed that S. divinorum was only weakly to
moderately correlated to the other four Salvia species based on the non-volatile compounds.
These results supported those observed using PCA and Student’s t-tests, which showed S.
divinorum was visually and statistically discriminated from the four other Salvia species. Thus,
it was possible to distinguish S. divinorum from the four other Salvia species based on the nonvolatile compounds analyzed and detected using LC-MS. With the GC-MS data (Chapter 3,
section 3.3.3), S. divinorum was distinguished from the four other Salvia species based on PCA
and Student’s t-tests; however, PPMC coefficients did not support this conclusion due to the
presence of the internal standard in all five species. Therefore, based on TICs, PPMC
coefficients, PCA, and Student’s t-tests, the chemical profiles generated using the LC-MS data
were more discriminatory than those generated using the GC-MS data.

4.4 Summary
S. divinorum and related Salvia species were extracted by rotary agitation for 16 hours
with acetonitrile and analyzed by LC-ESI-MS using positive ionization mode. Following
analysis, TICs were truncated, retention time aligned, and normalized prior to data analysis.
PPMC coefficients were calculated on TICs for replicate injections of the same extract and for
all replicates of each species. In general, moderate to strong correlations among injection and
extraction replicates indicated acceptable precision in the extraction, instrumental analysis, and
data pretreatment procedures.
121

S. divinorum was visually discriminated from the other four Salvia species based on
TICs. Using PCA, each Salvia species was visually discriminated from S. divinorum in the
scores plot. Student’s t-tests, calculated on the mean scores on both PCs, showed S. divinorum
was statistically differentiated from the four other Salvia species at the 99% confidence level.
PPMC coefficients were also calculated between S. divinorum and the other Salvia species.
Weak to moderate correlations were observed, which indicated chemical differences among S.
divinorum and the other four Salvia species. Using TICs, PCA, Student’s t-tests, and PPMC
coefficients, it was possible to distinguish S. divinorum from the other four Salvia species.
Therefore, based on TICs, PCA, Student’s t-tests, and PPMC coefficients, the chemical profiles
generated using the LC-MS data were more discriminatory than those generated using the GCMS data.

122

References

123

References
[1] Perrone A, Muzashvili T, Napolitano A, Skhirtladze A, Kemertelidze E, Pizza C,
Piacente S. Steroidal glycosides from the leaves of Ruscus colchicus: Isolation and structural
elucidation based on a preliminary liquid chromatography-electrospray ionization tandem
mass spectrometry profiling. Phytochemistry 2009; 70:2078–88.
[2] Medana C, Massolino C, Pazzi M, Baiocchi C. Determination of salvinorins and
divinatorins in Salvia divinorum leaves by liquid chromatography/multistage mass
spectrometry. Rapid Commun Mass Sp 2006; 20:131–6.
[3] Tsujikawa K, Kuwayama K, Miyaguchi H, Kanamori T, Iwata YT, Yoshida T, Inoue H.
Determination of salvinorin A and salvinorin B in Salvia divinorum-related products
circulated in Japan. Forensic Science International 2008; 180:105–9.
[4] Sana TR, Fischer S, Wohlgemuth G, Katrekar A, Jung K, Ronald PC, Fiehn O.
Metabolomic and transcriptomic analysis of the rice response to the bacterial blight pathogen
Xanthomonas oryzae pv. oryzae. Metabolomics 2010; 6:451–65.
[5] Fiehn O, Wohlgemuth G, Scholz M, Kind T, Lee DY, Lu Y, Moon S, Nikolau B. Quality
control for plant metabolomics: reporting MSI-complaint studies. The Plant Journal 2008;
53:691–704.
[6] Kennedy JH, Wiseman JM. Direct analysis of Salvia divinorum leaves for salvinorin A by
thin layer chromatography and desorption electrospray ionization multi-stage tandem mass
spectrometry. Rapid Commun Mass Sp 2010; 24:1305–11.
[7] Díaz-Cruz MS, López de Alda MJ, López R, Barceló D. Determination of estrogens and
progestogens by mass spectrometric techniques (GC/MS, LC/MS and LC/MS/MS). Journal of
Mass Spectrometry 2003; 38:917–23.
[8] Razboršek MI, Vončina DB, Doleček V, Vončina E. Determination of oleanolic, betulinic
and ursolic acid in Lamiaceae and mass spectral fragmentation of their trimethylsilylated
derivatives. Chromatographia 2008; 67:433–40.
[9] Shen D, Pan M, Wu Q, Park C, Juliani HR, Ho C et al. LC-MS method for the
simultaneous quantitation of the anti-inflammatory constituents in oregano (Origanum
Species). Journal of agricultural and food chemistry 2010; 58:7119–25.
[10] Fai YM, Tao CC. A review of presence of oleanolic acid in natural products. 2009;1–
271.
[11] Djarmati Z, Jankov RM, Djordjević A, Ribar B, Lazar D, Engel P. Carnosic acid 12methyl ether-y-lactone, a ferruginol-type diterpene from S. officinalis. Phytochemistry 1992;
31:1307–9.

124

[12] Chen H, Talaty NN, Takáts Z, Cooks RG. Desorption electrospray ionization mass
spectrometry for high-throughput analysis of pharmaceutical samples in the ambient
environment. Analytical Chemistry 2005; 77:6915–27.
[13] Lu Y, Foo LY. Polyphenolics of Salvia-a review. Phytochemistry 2002; 59:117–40.
[14] Ollanketo M, Peltoketo A, Hartonen K, Hiltunen R, Riekkola M. Extraction of sage (Salvia
officinalis L.) by pressurized hot water and conventional methods: antioxidant activity of the
extracts. European Food Research and Technology 2002; 215:158–63.
[15] Cuvelier ME, Berset C, Richard H. Antioxidant constituents in sage (Salvia officinalis).
Journal of Agricultural and Food Chemistry 1994; 42:665–9.
[16] Lamien-Meda A, Neil M, Lohwasser U, Börner A, Fraanz C, Novak J. Investigation of
antioxidant and rosmarinic acid variation in the sage collection of the genebank in Gatersleben.
Journal of Agricultural and Food Chemistry 2010; 58:3813–19.
[17] Cavero S, Jaime L, Martín-Álvarez PJ, Señoráns FJ, Reglero G, Ibañez E. In vitro
antioxidant analysis of supercritical fluid extracts from rosemary (Rosmarinus officinalis L.).
European Food Research and Technology 2005; 221:478–86.
[18] Walker JB, Sytsma KJ, Treutlein J, Wink M. Salvia (Lamiaceae) is not monophyletic:
implications for the systematics, radiation, and ecological specializations of Salvia and tribe
Mentheae. American Journal of Botany 2004; 91:1115–25.
[19] Bodnar Willard MA. Forensic analysis of Salvia divinorum and related Salvia species using
chemometric procedures. MS Thesis (2010).
[20] Tychonievich, J. Salvia clades. Personal Communication. 2009.
[21] Abreu ME, Müller M, Alegre L, Munné-Bosch S. Phenolic diterpene and α-tocopherol
contents in leaf extracts of 60 Salvia species. Journal of the Science of Food and Agriculture
2008; 88:2648–53.
[22] Schwarz K, Ternes W. Antioxidative constituents of Rosmarinus officinalis and Salvia
officinalis. Z Lebensm Unters Forsch 1992; 195:99–103.
[23] Adam C. Essential mathematics and statistics for forensic science. Hobken, NJ: WileyBlackwell 2010.

125

Chapter 5: Conclusions and Future Work
5.1 Gas Chromatography-Mass Spectrometry Analysis of Salvia divinorum and Related
Salvia Species
Salvia divinorum and four related Salvia species were extracted in dichloromethane and
analyzed using gas chromatography-mass spectrometry (GC-MS). All total ion chromatograms
(TICs) were then subjected to data pretreatment procedures (i.e. background correction,
smoothing, retention time alignment, and normalization) to reduce the non-chemical sources of
variance. Data analysis procedures (i.e. Pearson product moment correlation (PPMC)
coefficients, principal components analysis (PCA), and Student’s t-tests) were performed to
assess the discrimination of S. divinorum from the related species based on the volatile
compounds extracted from each. All five species contained common compounds; however, S.
divinorum was the only species containing salvinorins A, B, and D. PPMC coefficients
calculated between replicates of each species showed moderate to strong correlation, which
indicated acceptable precision in the extraction, instrumental analysis, and data pretreatment
procedures. PPMC coefficients also showed moderated to strong correlation between S.
divinorum and the other Salvia species. However, the internal standard added to all five species
increased the correlation between the species. Therefore, this issue needs to be further
investigated in the future. Despite some chemical similarities, S. divinorum was visually and
statistically discriminated from the four related Salvia species on the PCA scores plot, which
accounted for a total of 77% of the variance, using Student’s t-tests at the 99% confidence level.
The apparent contradiction between the PPMC and PCA results can be explained by considering
how PPMC coefficients and the PCA scores plot were calculated.

126
 

5.2 Liquid Chromatography-Mass Spectrometry Analysis of Salvia divinorum and Related
Salvia Species
S. divinorum and four related Salvia species were also extracted in acetonitrile and
analyzed using liquid chromatography-mass spectrometry (LC-MS) in positive ionization mode.
Once again, TICs were subjected to data pretreatment procedures (i.e. retention time alignment
and normalization) and the same data analysis procedures were performed. Common compounds
were shared among some species; however, S. divinorum was the only species containing
salvinorins A, B, C, D, and F. PPMC coefficients were calculated between replicates of each
species and showed moderate to strong correlation, which indicated acceptable precision in the
extraction, instrumental analysis, and data pretreatment procedures. PPMC coefficients
calculated between S. divinorum and the four other Salvia species showed weak to moderate
correlation. In addition, S. divinorum was visually and statistically discriminated from the four
related Salvia species on the PCA scores plot, which accounted for a total of 82% of the
variance, using Student’s t-tests at the 99% confidence level.

5.3 Comparison of Gas Chromatography-Mass Spectrometry and Liquid
Chromatography-Mass Spectrometry for the Analysis of Salvia divinorum and Related
Salvia Species
GC-MS is limited to the analysis of volatile compounds, while LC-MS can also be used
n

to analyze non-volatile compounds. In addition, LC-MS was previously shown to be more
sensitive than GC-MS; therefore, more non-volatile compounds are commonly detected when
n

using LC-MS and/or LC-MS [1]. For example, salvinorins A, B, C, D, and F were identified in
S. divinorum extracts analyzed using LC-MS, while only salvinorins A, B, and D were identified
in S. divinorum extracts analyzed using GC-MS. As previously stated, PPMC coefficients
calculated between replicates of each species analyzed using GC-MS and LC-MS both showed
127
 

moderate to strong correlation. This indicated acceptable precision in extraction and analysis for
both techniques. Mean PPMC coefficients calculated between S. divinorum and the other Salvia
species based on the volatile compounds were greater than 0.8, while mean PPMC coefficients
calculated based on more polar non-volatiles were below 0.8. However, the internal standard
increased the correlation between the species analyzed using GC-MS. Therefore, further
investigation of the use of the internal standard is needed in the future. Using PCA and Student’s
t-tests, S. divinorum was visually and statistically discrimination from the four related Salvia
species on the scores plot based on both the GC-MS and LC-MS data. Overall, the chemical
profiles generated using the LC-MS data were more discriminatory than those generated using
the GC-MS data based on TICs, PPMC coefficients, PCA, and Student’s t-tests.

5.4 Utility of Gas Chromatography-Mass Spectrometry and Liquid Chromatography-Mass
Spectrometry for the Analysis of Salvia divinorum
GC-MS, which is widely available in forensic laboratories, and LC-MS can be used to
identify S. divinorum based on the presence of salvinorin A [2]. Identification of S. divinorum in
this manner is possible since salvinorin A is only known to be found in S. divinorum thus far [3].
Identification of S. divinorum based on a full chemical fingerprint using LC-MS is
necessary in states were only S. divinorum is regulated and not salvinorin A, if salvinorin A is
later identified in other plants. However, the chemical composition of any plant material can vary
over time and based on the location where the plant is cultivated [4, 5]. Thus, the analyst must be
aware of this variability when identifying S. divinorum using a chemical fingerprint. Therefore,
both techniques are useful in analyzing S. divinorum.

128
 

5.5 Future Work
Further work can always improve previous research. In this research, S. divinorum and
four related Salvia species were extracted and analyzed to determine if S. divinorum could be
discriminated from four other Salvia species based on chemical fingerprints generated by GCMS and LC-MS. The extraction and analysis procedures in this research can be further
investigated. Firstly, different extraction solvents can be investigated. In this research,
dichloromethane and acetonitrile were used. It is polarity and temperature of the solvent that
determines which compounds are extracted from the plant material. Thus, other solvents or
mixture of solvents may extract more compounds from S. divinorum than observed in this
research. For example, the use of acetone, toluene, hexane, or mixtures of these solvents can be
investigated. Identifying more compounds in S. divinorum provides more information that can be
used to discriminate S. divinorum from other plant material.
Since GC-MS is limited to the analysis of volatile compounds, extracts containing nonvolatile compounds can be derivatized to modify compound volatility. Therefore, future work
may include the investigation of an extraction procedure with a dertivatization step so that more
compounds in S. divinorum can be detected using GC-MS. For example, the use of
bistrimethylsilyltrifluoroacetamide (BSTFA), and N-methyl-trimethylsilyltrifluoroacetamide
(MSTFA) as derivatizating reagents can be investigated. Thus, a more complete chemical
fingerprint of S. divinorum may be generated.
In this research, progesterone was used as the internal standard. However, as
progesterone is not stable in solution, the internal standard solution had to be prepared daily and
added to each extract. This caused variability in the abundance of the internal standard in the
resulting TICs, which affected subsequent data analysis procedures. Therefore, if an internal

129
 

standard is needed in future research, using a compound that is more stable over time is
beneficial. For example, the use 5-α-androstane, and cholesterol as an internal standard can be
investigated. Furthermore, the addition of the internal standard to the extracts analyzed by GCMS caused Salvia species to be more strongly correlated to one another based on PPMC
coefficients. Therefore, future work may include the investigation of the use of an internal
standard.
To further improve this research, different statistical procedures can be investigated. This
can include the investigation of soft supervised statistical procedures, such as soft independent
modeling by class analogy, which objectively identify an unknown extract based on how closely
it is associated to known extracts. The benefit of this type of statistical procedure is if the
unknown extract is not associated closely to any of the known extracts, an identification is not
forced. Therefore, this statistical procedure may be useful in a forensic laboratory.
In conclusion, both techniques are useful in analyzing S. divinorum and discriminating S.
divinorum from four other Salvia species. However, a more complete chemical fingerprint is
generated using LC-MS. Therefore, if an unknown plant material is submitted to a forensic
laboratory and can not be identified by GC-MS, an alternative procedure for identifying the plant
material based on several compounds using LC-MS should be considered.

 
 
 
 
 

130
 

 
 
 
 
 
 
 
 
 
 

References

131
 

References
[1] Medana C, Massolino C, Pazzi M, Baiocchi C. Determination of salvinorins and divinatorins
in Salvia divinorum leaves by liquid chromatography/multistage mass spectrometry. Rapid
Commun Mass Sp 2006; 20:131–6.
[2] Watson JT, Sparkman OD. Introduction to mass spectrometry: instrumentation, applications
and strategies for data interpretation. 4th ed. West Sussex, England: John Wiley & Sons Ltd,
2007.
[3] Giroud C, Felber F, Augsburger M, Horisberger B, Rivier L, Mangin P. Salvia divinorum: an
hallucinogenic mint which might become a new recreational drug in Switzerland. Forensic Sci
Int 2000; 112:143–50.
[4] Siebert, DJ. Localization of salvinorin A and related compounds in glandular trichomes of the
psychoactive sage, Salvia divinorum. Annals of Botany 2004; 93:763–71.
[5] Carrer RP, Vanderlinde R, Dutra S, Marcon A, Echeverrigaray S. Esstential oil varation
among Brazilian accessions of Salvia guaranitica L.. Flavour Fragr. J. 2007; 22:430–4.

132
 

 
 
 
 
 
 
 
 
 
 

APPENDICES
 
 
 
 
 
 
 
 
 
 
 
 
 
 

133
 

 
 

APPENDIX A

PROVISIONAL IDENTIFICATION OF COMPOUNDS IN THE FIVE SALVIA SPECIES
ANALYZED USING GAS CHROMATOGRAPHY-MASS SPECTROMETRY

134
 

Molecular Mass
(m/z)

23.24

380

Provisional
Identification of
Compound and/or
Compound Class*

Approximate
Retention Time
(minutes)

1

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table A.1: Provisional Identification of Compounds in S. divinorum Analyzed Using Gas
Chromatography-Mass Spectrometry

71, 57, 97, 85, 83, 84,
Heptacosane
113, 194, 127, 154
2
24.03
410
69, 81, 95, 77, 94, 137,
Squalene
387, 295, 80, 154
3
24.47
314
124, 314, 272, 229, 91,
Progesterone
79, 105, 107, 93, 133
4
24.61
408
71, 57, 85, 99, 83, 177,
Nonacosane
55, 118, 125, 141
5
24.82
**
69, 204, 96, 81, 53, 107,
Acetate
91, 164, 109, 97
6
25.94
436
71, 57, 85, 56, 96, 99, 97, Hentriacontane
443, 295, 127
7
26.17
390
94, 291, 237, 166, 304,
Salvinorin B
356, 476, 305, 107, 77
8
26.89
432
94, 273, 432, 121, 95,
Salvinorin A
107, 166, 220, 91, 404
9
27.17
478
71, 57, 85, 55, 97, 99,
Tetratriacontane
113, 56, 70, 69
10
27.29
432
94, 179, 372, 105, 97,
Salvinorin D
206, 400, 77, 237, 383
11
27.44
414
397, 91, 268, 414, 145,
Sitosterol
273, 95, 60, 81, 159
12
28.35
492
57, 71, 85, 206, 55, 99,
Pentatriacontane
97, 463, 134, 88
*Provisional identification based on library searches and literature sources
**Unknown molecular mass 
 
 
 
 
 

135
 

Molecular Mass
(m/z)

24.03

410

Provisional
Identification of
Compound and/or
Compound Class*

Approximate
Retention Time
(minutes)

2

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table A.2: Provisional Identification of Compounds in S. guaranitica Analyzed Using Gas
Chromatography-Mass Spectrometry

69, 81, 105, 95, 342, 93,
Squalene
68, 149, 77, 68
3
24.43
314
124, 314, 272, 229, 91, 79, Progesterone
105, 107, 133, 77
6
25.94
436
57, 71, 85, 99, 55, 113, 69, Hentriacontane
83, 97, 127
9
27.18
478
57, 71, 85, 99, 55, 83, 69, Tetratriacontane
111, 72, 141
11
27.46
414
414, 396, 329, 303, 399,
Sitosterol
381, 159, 95, 107, 213
13
9.91
134
105, 77, 134, 76, 51, 106, Phthalide
63, 50, 104, 89
14
11.75
204
161, 91, 105, 204, 79, 120, β-Cubebene
77, 93, 81, 162
15
25.52
340
143, 340, 202, 128, 129,
Ester
95, 142, 157, 203, 141
16
25.65
338
160, 141, 115, 145, 129,
Alkene
128, 91, 95, 94, 159
17
26.18
430
430, 165, 431, 164, 205,
Vitamin E
166, 432, 55, 121, 57
18
26.45
338
141, 94, 95, 81, 357, 145, Dione
338, 128, 142, 222
19
26.66
340
340, 95, 94, 133, 82, 91,
Dione
128, 189, 77, 105
20
26.83
340
135, 94, 121, 81, 91, 340, Nitrolepidine
95, 77, 129, 105
21
27.30
399
164, 94, 81, 95, 67, 329,
Benzenesulfonamide
77, 121, 161, 491
22
27.80
426
218, 203, 219, 105, 95, 94, β-Amyrin
121, 189, 81, 204
23
28.09
426
218, 145, 136, 122, 161,
α-Amyrin
55, 203, 109, 189, 426
*Provisional identification based on library searches and literature sources

136
 

Molecular Mass
(m/z)

24.45

314

Provisional
Identification of
Compound and/or
Compound Class*

Approximate
Retention Time
(minutes)

3

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table A.3: Provisional Identification of Compounds in S. nemorosa Analyzed Using Gas
Chromatography-Mass Spectrometry

124, 314, 272, 229, Progesterone
91, 79, 105, 93, 77,
107
4
24.63
408
57, 71, 85, 55, 97,
Nonacosane
83, 69, 56, 41, 86
6
25.95
436
57, 71, 85, 55, 69,
Hentriacontane
83, 99, 97, 56, 70
9
27.19
478
57, 7, 85, 55, 97,
Tetratriacontane
99, 83, 69, 56, 113
11
27.47
414
414, 107, 93, 97,
Sitosterol
396, 303, 255, 55,
329, 145
14
11.76
204
161, 105, 204, 79,
β-Cubebene
73, 120, 69, 93, 80,
91
17
26.19
430
430, 165, 431, 164, Vitamin E
166, 205, 57, 121,
69, 203
24
10.98
204
93, 69, 79, 9, 95,
β-Caryophyllene
77, 55, 161, 175, 94
25
15.68
123
95, 55, 123, 68, 82, Alkene
403, 221, 488, 98,
57
26
20.57
213
72, 69, 59, 96, 97,
Amide
126, 82, 165, 56,
114
27
26.25
328
313, 328, 153, 285, Phenyl
299, 69, 314, 181,
268, 329
28
26.58
478
57, 71, 85, 97, 99,
11-n-Decyltetracosane
55, 192, 56, 83, 69
29
28.37
506
57, 71, 96, 97, 85,
Hextriacontane
206, 69, 179, 339,
82
*Provisional identification based on library searches and literature sources
137
 

Molecular
Mass (m/z)

23.24

380

Provisional
Identification
of Compound
and/or
Compound
Class*

Approximate
Retention
Time
(minutes)

1

Major
Fragment Ions
(m/z)
(Order of
decreasing
abundance)

Compound
Number

Table A.4: Provisional Identification of Compounds in S. officinalis Analyzed Using Gas
Chromatography-Mass Spectrometry

57, 71, 85, 99, 55, 113,
Heptacosane
84, 141, 69, 83
2
24.07
410
69, 81, 67, 68, 287, 122, Squalene
255, 123, 70, 94
3
24.47
314
124, 314, 272, 229, 91,
Progesterone
79, 105, 107, 133, 93
4
24.65
408
71, 57, 85, 97, 99, 113,
Nonacosane
124, 72, 182, 169
6
25.97
436
57, 71, 85, 97, 99, 113,
Hentriacontane
69, 55, 83, 56
9
27.21
478
57, 71, 85, 99, 97, 69,
Tetratriacontane
55, 83, 113, 127
11
27.48
414
414, 213, 396, 399, 303, Sitosterol
159, 97, 145, 57, 95
17
26.21
430
165, 430, 164, 431, 166, Vitamin E
205, 432, 57, 121, 206
23
28.11
426
97, 218, 161, 87, 189,
α-Amyrin
69, 203, 95, 81, 217
25
15.70
123
95, 68, 123, 55, 97, 82,
Alkene
69, 57, 124, 83
29
28.40
506
57, 71, 85, 55, 69, 99,
Hexatriacontane
83, 113, 70, 97
30
3.95
136
93, 91, 92, 77, 79, 121,
α-Pinene
94, 136, 80, 105
31
4.19
136
93, 121, 79, 107, 91, 67, Camphene
77, 94, 136, 95
32
4.56
136
93, 69, 91, 77, 79, 80,
β-Pinene
92, 67, 94, 121
33
4.60
136
93, 69, 208, 92, 94, 77,
β-Myrcene
67, 53, 70, 79
34
5.32
154
81, 154, 108, 139, 111,
Eucalyptol
84, 71, 93, 69, 83
35
6.43
152
110, 81, 109, 68, 95, 67, Thujone
69, 82, 79, 55
36
6.60
152
110, 81, 95, 109, 67, 69, Thujone
68, 70, 55, 79
*Provisional identification based on library searches and literature sources
138
 

Molecular
Mass (m/z)

7.10

152

Provisional
Identification
of Compound
and/or
Compound
Class*

Approximate
Retention
Time (minutes)

37

Major
Fragment Ions
(m/z)
(Order of
decreasing
abundance)

Compound
Number

Table A.4 (cont’d)

95, 81, 108, 152, 109,
Camphor
69, 83, 55, 67, 110
38
7.45
154
95, 110, 139, 93, 55, 67, Borneol
79, 111, 84, 69
39
9.06
196
95, 93, 121, 136, 92,
Bornyl Acetate and/or
108, 80, 91, 110, 67
Isoborynyl Acetate
40
9.93
168
153, 108, 107, 93, 109,
Bicyclic Compound
111, 69, 55, 99, 79
41
11.00
164
93, 69, 91, 79, 120, 107, Cyclopropane Compound
105, 106, 67, 95
42
11.47
204
93, 80, 121, 92, 91, 107, α-Caryophyllene
94, 79, 77, 109
43
13.20
222
109, 161, 93, 105, 69,
Epiglobulol
107, 91, 122, 204, 81
44
13.39
220
109, 138, 67, 93, 123,
Diene
95, 68, 55, 81, 82
45
18.05
290
137, 95, 81, 257, 93, 69, Epimanool
123, 107, 109, 79
46
21.09
330
286, 230, 215, 243, 218, Isocarnosol
271, 204, 217, 287, 203
47
21.37
286
271, 286, 272, 128, 287, Phenyl
215, 229, 141, 228, 135
48
21.75
286
286, 215, 271, 204, 287, Dione
202, 187, 141, 216, 272
49
26.44
422
57, 71, 85, 56, 97, 55,
Squalane
99, 69, 70, 83
50
26.98
422
57, 71, 85, 99, 97, 55,
Triacontane
69, 83, 113, 56
51
27.66
450
57, 71, 85, 56, 97, 55,
Dotriacontane
83, 99, 69, 113
52
27.81
478
57, 71, 85, 83, 69, 99,
9-n-Octylhexacosane
55, 113, 70, 97
53
29.72
456
203, 218, 97, 415, 282,
Oleanolic Acid
133, 95, 149, 135, 119
*Provisional identification based on library searches and literature sources

139
 

Molecular Mass
(m/z)

24.06

410

Provisional
Identification of
Compound and/or
Compound Class*

Approximate
Retention Time
(minutes)

2

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table A.5: Provisional Identification of Compounds in S. splendens Analyzed Using Gas
Chromatography-Mass Spectrometry

69, 81, 95, 136, 121, Squalene
68, 137, 123, 93, 67
3
24.46
314
124, 229, 272, 314,
Progesterone
91, 79, 105, 93, 107,
95
4
24.63
408
57, 71, 85, 99, 69,
Nonacosane
55, 113, 56, 83, 111
6
25.96
436
57, 71, 85, 97, 99,
Hentriacontane
69, 55, 83, 113, 56
9
27.20
478
57, 71, 85, 55, 99,
Tetratriacontane
83, 69, 97, 113, 70
11
27.48
414
414, 81, 107, 105,
Sitosterol
145, 95, 57, 55, 303,
93
17
26.20
430
165, 430, 164, 431,
Vitamin E
166, 205, 57, 55, 69,
97
23
28.12
426
218, 189, 123, 107,
α-Amyrin
203, 136, 219, 122,
79, 109
25
15.70
123
95, 68, 82, 123, 57,
Alkene
55, 97, 71, 69, 83
29
28.38
506
57, 71, 85, 97, 55,
Hextriacontane
99, 69, 83, 56, 113
53
29.71
456
207, 203, 73, 133,
Oleanolic Acid
59, 179, 206, 125,
208, 189
54
26.42
342
94, 91, 95, 117, 342, 8-Episalviarin
105, 121, 81, 77, 145
55
27.03
406
189, 91, 110, 340,
Cardenolide
97, 129, 95, 81, 143,
105
*Provisional identification based on library searches and literature sources

140
 

APPENDIX B

PROVISIONAL IDENTIFICATION OF COMPOUNDS IN THE FIVE SALVIA SPECIES
ANALYZED USING LIQUID CHROMATOGRAPHY-MASS SPECTROMETRY

141
 

Compound Number

Approximate
Retention Time
(minutes)

Molecular Mass
(m/z)
Molecular Mass +
Hydrogen (m/z)
Adduct Ions (m/z)

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Major Ions of
Enhanced Product
Ion 1 (m/z)

Major Ions of
Enhanced Product
Ion 2 (m/z)

Provisional
Identification of
Compound and/or
Compound Class*

Table B.1: Provisional Identification of Compounds in S. divinorum Analyzed Using Liquid
Chromatography-Mass Spectrometry

1

9.53

390

391

413,
423,
803

373, 391, 374,
355, 259, 413,
327, 392, 341,
414

(of 391) 145,
105, 119, 107,
131, 91, 173,
117, 143, 121

Salvinorin
B

2**

10.13

406

407

424,
835

424, 429, 341,
373, 407, 329,
413, 389, 445,
214

(of 424) 157,
293, 309, 355,
183, 209, 179,
195, 185, 265

3**

10.32

378

379

396,
401

311, 214, 371,
401, 496, 343,
312, 361, 279,
158

(of 524) 157,
309, 355, 185,
156, 291, 337,
183, 235

4

10.73

432

433

455

373, 341, 391,
214, 374. 323,
455, 279, 413,
355

(of 373) 121,
341, 145, 91,
105, 231, 107,
213, 223, 129

5

11.20

374

375

397,
771

325, 357, 326,
397, 214, 358,
331, 279, 87,
158

(of 325) 325,
105, 145, 179,
297, 91, 223,
117, 143, 119

(of 366)
142, 193,
423, 199,
91, 105,
119, 145,
213, 195
(of 407)
191, 149,
105, 107,
133, 147,
173, 223,
90, 104
(of 373)
191, 129,
209, 323,
401, 91,
107, 115,
121, 127
(of 341)
105, 145,
91, 109,
117, 173,
103, 121,
147, 157
(of 331)
201, 279,
105, 128,
149, 93,
105, 155,
173, 397

*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time

142
 

Terpenoid

Terpenoid

Salvinorin
D

Salvinorin
F

432

433

455,
887

373, 374,
455, 341,
313, 355,
314, 375,
295, 456

(of 331)
(of 408)
295, 135,
91, 107,
121, 107,
121, 145,
91, 105,
187, 199,
131, 187,
135, 131,
277, 145
211, 221
7
12.69
474 475 497, 415, 497,
(of 279)
(of 499)
971
475, 305,
155, 156,
497, 219,
261, 416,
157, 142,
95, 145,
355, 279,
105, 131,
147, 189,
214, 323
145, 129,
194, 219,
121, 141
498, 91
8
18.59
264 265 282, 282, 304,
(of 313)
(of 279)
304
283, 265,
93, 97,
282, 201,
247, 563,
107, 95,
247, 107,
305, 284,
135, 121,
265, 97,
266, 248
111, 100,
177, 283,
91, 282
304, 149
9** 20.83
426 427 444, 409, 410,
(of 429)
(of 447)
449
449, 259,
409, 107,
335, 149,
327, 271,
95, 109,
105, 613,
285, 203,
119, 105,
91, 119,
450, 257
121, 133,
95, 595,
159, 91
93, 107
*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time

143
 

Provisional
Identification of
Compound and/or
Compound Class*

Major Ions of
Enhanced Product
Ion 2 (m/z)

12.09

Major Ions of
Enhanced Product
Ion 1 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)
Molecular Mass +
Hydrogen (m/z)
Adduct Ions (m/z)

6

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table B.1 (cont’d)

Salvinorin A

Salvinorin C

Sesquiterpenoid

Unidentified
Compound

Adduct Ions (m/z)

264

265

282,
304

282, 304,
283, 265,
247, 563,
305, 284,
348, 266

10

9.56

340

341

363,
703

341, 363,
295, 342,
323, 364,
296, 324,
249, 343

11

10.16

398

399

421,
819

339, 399,
321, 421,
340, 293,
381, 322,
400, 357

12**

10.36

340,
398

341,
399

421,
703

341, 295,
342, 323,
363, 421,
296, 324,
381, 343

13

12.36

398

399

421,
819

339, 399,
421, 321,
340, 245,
293, 357,
400, 422

Provisional
Identification of
Compound and/or
Compound Class*

Molecular Mass +
Hydrogen (m/z)

18.77

Major Ions of
Enhanced Product
Ion 2 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)

8

Major Ions of
Enhanced Product
Ion 1 (m/z)

Compound Number

Major Fragment Ions
(m/z)
(Order of decreasing
abundance)

Table B.2: Provisional Identification of Compounds in S. guaranitica Analyzed Using Liquid
Chromatography-Mass Spectrometry

(of 282)
93, 91,
95, 107,
97, 121,
114, 135,
86, 95
(of 357)
95, 161,
165, 179,
141, 129,
115, 155,
128, 178
(of 343)
178, 165,
179, 191,
128, 203,
207, 202,
141, 205
(of 323)
237, 165,
141, 295,
169, 155,
179, 154,
129, 167
(of 345)
128, 155,
141, 129,
142, 178,
153, 179,
165, 115

(of 331)
270, 107,
298, 316,
331, 91,
93, 109,
301, 92
(of 293)
165, 128,
171, 215,
178, 117,
128, 131,
153, 95
(of 279)
91, 129,
178, 128,
131, 179,
219, 115,
165, 141
(of 295)
165, 155,
153, 179,
167, 169,
154, 178,
202, 203
(of 339)
179, 142,
178, 189,
165, 155,
178, 192,
115, 141

Sesquiterpenoid

*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time
144
 

Diterpenoid

Terpenoid

Terpenoid

Terpenoid

457

479,
935

479, 439, 191,
480, 440, 411,
203, 393, 501,
331

15

21.46

430

431

453

431, 430, 429,
432, 445, 331,
301, 397, 433,
446

16

21.74

***

***

***

429, 522, 430,
301, 331, 523,
239, 185, 397,
141

Provisional
Identification of
Compound and/or
Compound Class*

Adduct Ions (m/z)

456

Major Ions of
Enhanced Product
Ion 2 (m/z)

Molecular Mass +
Hydrogen (m/z)

18.22

Major Ions of
Enhanced Product
Ion 1 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)

14**

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table B.2 (cont’d)

(of 302)
479, 95,
91, 93,
105, 107,
109, 121,
134, 435
(of 331)
165, 137,
109, 119,
166, 121,
431, 122,
107, 205
(of 429)
637, 165,
107, 105,
91, 93,
105, 145,
187, 121

(of 280)
203, 119,
105, 133,
95, 107,
93, 121,
147, 109
(of 495)
165, 137,
139, 494,
107, 91,
109, 93,
105, 149
(of 409)
522, 697,
298, 270,
523, 524,
301, 93,
525, 95

Oleanolic
and/or
Ursolic
Acid

*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time
***Unknown molecular mass, molecular mass + hydrogen, and adduct ions

145
 

Vitamin E

Unidentified
Compound

264

265

282,
304

282, 283,
304, 563,
247, 265,
564, 284,
305, 248

(of 282)
(of 191)
93, 91,
91, 93,
107, 95,
97, 135,
135, 97,
95, 109,
121, 111,
121, 304,
100, 114
83, 94
14** 18.19
456 457 479, 479, 439,
(of 245)
(of 263)
935
480, 191,
119, 203,
119, 107,
440, 411,
479, 105,
133, 203,
457, 481,
133, 107,
105, 95,
203, 936
95, 93,
121, 109,
191, 147
191, 93
15
21.33
430 431 453
431, 430,
(of 329)
(of 730)
432, 429,
165, 137,
599, 237,
445, 895,
119, 166,
597, 595,
411, 329,
431, 268,
895, 268,
896, 413
109, 167,
317, 237
136
17
20.91
*** *** ***
429, 430,
(of 469)
(of 869)
447, 469,
165, 149,
573, 469,
397, 329,
177, 137,
149, 595,
960, 954,
163, 429,
163, 317,
448, 470
191, 205,
107, 165,
119, 135
261, 119
*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time
***Unknown molecular mass, molecular mass + hydrogen, and adduct ions

146
 

Provisional
Identification of
Compound and/or
Compound Class*

Major Ions of
Enhanced Product
Ion 2 (m/z)

18.63

Major Ions of
Enhanced Product
Ion 1 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)
Molecular Mass +
Hydrogen (m/z)
Adduct Ions (m/z)

8

Major Fragment Ions
(m/z)
(Order of decreasing
abundance)

Compound Number

Table B.3: Provisional Identification of Compounds in S. nemorosa Analyzed Using Liquid
Chromatography-Mass Spectrometry

Sesquiterpenoid

Oleanolic
and/or Ursolic
Acid

Vitamin E

Unidentified
Compound

Compound Number

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)
Molecular Mass +
Hydrogen (m/z)

Adduct Ions (m/z)

Major Fragment Ions
(m/z)
(Order of decreasing
abundance)

Major Ions of
Enhanced Product
Ion 1 (m/z)

Major Ions of
Enhanced Product
Ion 2 (m/z)

Provisional
Identification of
Compound and/or
Compound Class*

Table B.4: Provisional Identification of Compounds in S. officinalis Analyzed Using Liquid
Chromatography-Mass Spectrometry

8

18.62

264

265

282,
304

282, 283,
304, 265,
247, 563,
284, 305,
266, 248

(of 282) 93,
91, 121, 95,
107, 97,
111, 135,
240, 109

Sesquiterpenoid

14**

18.12

456

457

479,
935

439, 479,
191, 440,
285, 411,
480, 457,
301, 203

(of 304)
91, 107,
93, 95,
304, 313,
133, 92,
105, 109
(of 285)
256, 91,
93, 107,
103, 285,
95, 121,
105, 131
(of 569)
285, 283,
527, 191,
215, 243,
117, 569,
115, 131
(of 301)
286, 153,
271, 105,
119, 131,
140, 153,
168, 301
(of 301)
103, 217,
149, 191,
115, 131,
128, 129,
205, 145

(of 439)
203, 119,
107, 105,
95, 133,
147, 121,
189, 109
15
21.30
430 431
453 431, 430,
(of 589)
432, 429,
165, 599,
285, 301,
595, 137,
205, 433,
895, 317,
331, 445
109, 119,
261, 107
18
10.24
300 301
323 301, 302,
(of 271)
323, 303,
286, 301,
214, 297,
168, 140,
324, 271,
287, 121,
158, 279
258, 285,
112, 271
19** 15.02
330 331, 354, 287, 205,
(of 331)
,
333
355, 331, 288,
115, 117,
332
683, 245, 285,
215, 103,
685
191, 355,
197, 145,
333, 353
285, 267,
225, 131
*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time

147
 

Oleanolic
and/or Ursolic
Acid

Vitamin E

3-OMethylkaempfe
rol

Carnosol and
Carnosic Acid

347

369,
715

301, 219,
205, 302,
220, 369,
206, 347,
231, 303

21

20.00

***

***

***

285, 551,
331, 287,
301, 273,
569, 205,
191, 552

Provisional
Identification of
Compound and/or
Compound Class*

346

Major Ions of
Enhanced Product
Ion 2 (m/z)

Adduct Ions (m/z)

15.96

Major Ions of
Enhanced Product
Ion 1 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)
Molecular Mass +
Hydrogen (m/z)

20**

Major Fragment
Ions (m/z)
(Order of
decreasing
abundance)

Compound Number

Table B.4 (cont’d)

(of 219)
231, 148,
205, 103,
115, 131,
128, 129,
219, 162
(of 285)
119, 105,
133, 145,
93, 91,
107, 121,
143, 173

(of 205)
103, 102,
148, 130,
128, 115,
369, 323,
91, 147
(of 331)
197, 215,
285, 267,
169, 225,
243, 165,
157, 191

Methyl
Carnosate,
Rosmanol
and/or
Epirosmanol

*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time
***Unknown molecular mass, molecular mass + hydrogen, and adduct ions

148
 

Unidentified
Compound

282,
304

282, 304,
283, 247,
265, 563,
305, 284,
248, 266

(of 479)
(of 443)
93, 97, 95, 331, 107,
91, 107,
313, 301,
121, 111,
91, 93,
109, 100,
271, 331,
114
94, 95
9
21.03
426
427
444, 409, 410,
(of 409)
(of 285)
449 449, 427,
107, 119,
105, 107,
444, 285,
105, 409,
613, 91,
259, 450,
95, 109,
119, 109,
327, 411
121, 91,
93, 151,
93, 133
159, 189
14** 18.14
456
457
479, 439, 479,
(of 302)
(of 245)
935 191, 440,
119, 107,
119, 107,
480, 411,
133, 105,
479, 105,
457, 203,
203, 95,
95, 121,
393, 458
121, 91,
93, 133,
109, 191
147, 163
15
21.32
430
431
453 431, 430,
(of 430)
(of 445)
432, 429,
165, 430,
119, 147,
425, 445,
164, 137,
175, 107,
433, 447,
121, 205,
123, 137,
407, 426
136, 119,
139, 163,
109, 91
445, 95
22
9.85
400
401
423, 341, 339,
(of 267)
(of 303)
823 401, 423,
341, 129,
179, 109,
383, 342,
117, 179,
157, 341,
399, 357,
105, 115,
105, 129,
340, 421
157, 165,
143, 178,
128, 142
91, 95
*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time

149
 

Provisional
Identification of
Compound and/or
Compound Class*

Adduct Ions (m/z)

265

264

MS2 Major Ions of
EPI 2 (m/z)

Molecular Mass +
Hydrogen (m/z)

18.61

Major Ions of
Enhanced Product
Ion 1 (m/z)

Approximate
Retention Time
(minutes)
Molecular Mass
(m/z)

8

Major Fragment Ions
(m/z)
(Order of decreasing
abundance)

Compound Number

Table B.5: Provisional Identification of Compounds in S. splendens Analyzed Using Liquid
Chromatography-Mass Spectrometry

Sesquiterpenoid

Unidentified
Compound

Oleanolic/
Ursolic Acid

Vitamin E

Diterpenoid

Molecular Mass +
Hydrogen (m/z)

Adduct Ions (m/z)

342

343

365,
707

343, 365,
344, 325,
297, 251,
231, 279,
707, 326

24

10.21

362

363

***

305, 363,
287, 323,
275, 364,
306, 269,
288, 231

25**

12.10

406,
548

407,
549

572,
587

311, 566,
293, 549,
312, 329,
371, 571,
345, 265

Provisional
Identification of
Compound and/or
Compound Class*

Molecular Mass
(m/z)

10.04

MS2 Major Ions of
EPI 2 (m/z)

Approximate
Retention Time
(minutes)

23

Major Ions of
Enhanced Product
Ion 1 (m/z)

Compound Number

Major Fragment Ions
(m/z)
(Order of decreasing
abundance)

Table B.5 (cont’d)

(of 325)
117, 129,
157, 115,
131, 128,
142, 91,
143, 141
(of 325)
363, 91,
105, 107,
95, 119,
117, 129,
121, 93
(of 345)
284, 312,
345, 330,
91, 115,
131, 117,
128, 105

(of 365)
192, 157,
165, 166,
167, 179,
199, 203,
207, 217
(of 365)
105, 91,
119, 93, 95,
131, 133,
147, 117,
145
(of 311)
217, 171,
107, 131,
145, 311,
199, 189,
265, 229

Diterpenoid

*Provisional identification based on literature sources
**Possibly two or more compounds coeluting at this retention time
***Unknown adduct ions

150
 

Diterpenoid

Phenolic
Methyl Ether
and
Unidentified
Compound