identification and isolation of drugs in complex...
TRANSCRIPT
Identification and Isolation of Drugs in Complex
Samples
By
Niki Kim Burns
B.ForensicSc (Hons)
Submitted in fulfilment of the requirements for the degree of
Doctor of Philosophy
Deakin University
May, 2017
IV
"All our dreams can come true, if we have the courage to pursue them."
- Walt Disney
V
TABLE OF CONTENTS
TABLE OF CONTENTS ........................................................................... V
ABSTRACT .............................................................................................. IX
ACKNOWLEDGEMENTS .................................................................... XII
PUBLICATIONS ................................................................................... XIV
ABBREVIATIONS ................................................................................. XV
CHAPTER 1: ............................................................................................... 1
INTRODUCTION ....................................................................................... 1
Complex samples in context .................................................................... 2
Opiates ................................................................................................. 2
Synthetic cannabinoids ........................................................................ 6
Separation and Detection methods ........................................................ 10
High Performance Liquid Chromatography (HPLC) ......................... 11
Two-Dimensional High Performance Liquid Chromatography (2D-
HPLC) ............................................................................................................ 12
Other applied separation techniques .................................................. 13
Chemiluminescence ........................................................................... 14
Electrochemiluminescence (ECL) ..................................................... 16
Mass Spectrometry (MS) ................................................................... 17
Nuclear Magnetic Resonance Spectroscopy (NMR) ......................... 18
Data Analysis ......................................................................................... 18
Project aims ........................................................................................... 19
CHAPTER 2: ............................................................................................. 21
BLIND COLUMN SELECTION PROTOCOL FOR TWO-DIMENSIONAL HIGH PERFORMANCE LIQUID CHROMATOGRAPHY... 21
Chapter Overview .................................................................................. 22
Introduction ............................................................................................ 23
Experimental .......................................................................................... 28
VI
Chemicals and samples ...................................................................... 28
Liquid Chromatography Mass Spectrometry (LC/MS) ..................... 28
Two-Dimensional High Performance Liquid Chromatography ........ 30
Data analysis ...................................................................................... 30
Results and discussion ........................................................................... 32
Conclusions ............................................................................................ 51
CHAPTER 3: ............................................................................................. 52
OPIATE PROCESS STREAM ANALYSIS ............................................ 52
Chapter overview ................................................................................... 53
Introduction ............................................................................................ 54
Experimental .......................................................................................... 56
Chemicals and samples ...................................................................... 56
Two-Dimensional High Performance Liquid Chromatography ........ 56
Principal Components Analysis (PCA) .............................................. 56
Liquid Chromatography/ Mass Spectrometry (LC/MS) .................... 58
Results and discussion ........................................................................... 60
Principal Components Analysis (PCA) .............................................. 63
Mass Spectrometry ............................................................................. 72
Conclusions ............................................................................................ 76
CHAPTER 4: ............................................................................................. 77
APPLICATION OF PLATFORM TECHNOLOGY TO MONITOR COMPONENTS OF INTEREST .......................................................................... 77
Chapter overview ................................................................................... 78
Introduction ............................................................................................ 79
Molecule of Interest 1 (285 amu) ....................................................... 80
Molecule of Interest 2 (622 amu) ....................................................... 80
Molecule of Interest 3 (297 amu) ....................................................... 81
Experimental .......................................................................................... 84
Case studies ........................................................................................ 84
VII
Results and discussion ........................................................................... 85
Molecules of interest .......................................................................... 85
Case 1: G1527-027N, G1527-027O and G1527-027G. ..................... 92
Case 2: G1527-027L and G1527-027M........................................... 100
Case 3: G1528-028C, G1527-028D and G1527-028H .................... 104
Conclusions .......................................................................................... 110
CHAPTER 5: ........................................................................................... 111
IDENTIFICATION AND SELECTIVE DETECTION OF SYNTHETIC CANNABINOIDS ............................................................................................... 111
Chapter overview ................................................................................. 112
Introduction .......................................................................................... 113
Solid Phase Extraction ..................................................................... 114
Experimental ........................................................................................ 115
Chemicals, samples and standards ................................................... 115
High Performance Liquid Chromatography (HPLC) ....................... 116
Chemiluminescence detection .......................................................... 117
Electrospray mass spectrometry ....................................................... 117
Electrochemiluminescence (ECL) and Cyclic Voltammetry (CV) .. 118
Solid Phase Extraction (SPE) ........................................................... 118
Results and discussion ......................................................................... 120
Synthetic cannabinoid extraction ..................................................... 120
Identification .................................................................................... 122
Chemiluminescence detection .......................................................... 125
Electrochemiluminescence (ECL) and Cyclic Voltammetry (CV) .. 134
Solid Phase Extraction ..................................................................... 139
Conclusions .......................................................................................... 144
CHAPTER 6: ........................................................................................... 145
TARGETED NMR SPECTROSCOPY FOR SELECTIVE DETERMINATION OF SYNTHETIC CANNABINOIDS ............................... 145
VIII
Chapter overview ................................................................................. 146
Introduction .......................................................................................... 147
Experimental ........................................................................................ 150
Samples ............................................................................................ 150
Instrumentation and Experiments .................................................... 150
Results and discussion ......................................................................... 151
Conclusions .......................................................................................... 160
Future Work ............................................................................................ 161
REFERENCES ........................................................................................ 163
IX
ABSTRACT
Complex samples are commonly problematic in chemistry and pose many
challenges in both forensic science and natural product extractions. As
developments in analytical instrumentation affords increases in sensitivity,
complex samples become more of an issue as previously undetectable levels of
chemical components can now be identified. There is a demand therefore driving a
need for improved protocols to deal with the comprehensive data generated from
this superior technology. This allows the analysts to observe previously
undetectable analytes, while limiting the impact of the complexity of data analysis.
As such, investigations into new methodologies are warranted for the isolation and
identification of drugs in complex samples in both forensic and pharmaceutical
applications.
Synthetic cannabinoids are a relatively new class of illicit substance and due
to the large numbers of emerging compounds, these require attention to aid law
enforcement agencies. Consequently, in this thesis these compounds have been
isolated and identified from within complex herbal substrates using a range of
optimised novel sample preparations and analytical instruments. Likewise,
methods for mapping the Sun Pharmaceutical Industries Australia opiate
processing streams were developed, and peaks of interest between stages of the
process were identified to meet a genuine industrial problem for this multi-national
company. Changes in the process and the subsequent impact of the molecules
present was then investigated with the aid of mass spectrometry. This was carried
out to inform the company of origins of impurities and major differences occurring
in the extraction line so that better response times to process adjustments can be
made.
Two-dimensional high performance liquid chromatography (2D-HPLC)
was utilised for the separation of opiate extractions from opium poppy (Papaver
Somniferum). Multidimensional chromatographic separations require the selection
of two orthogonal columns; this is a labour intensive and time consuming process,
which in many cases is an entirely trial-and-error approach. This thesis introduces
a novel blind optimisation method for column selection of a black box of
constituent components. A data processing pipeline, created in the open source
X
application OpenMS®, was developed to map the components of equal mass within
the mixture across a library of HPLC columns; 2D-HPLC separation space
utilisation was compared by measuring the fractional surface coverage, fcoverage. It
was found that for a test mixture from an opium poppy (Papaver somniferum)
extract, the combination of diphenyl and C18 stationary phases provided a
predicted fcoverage of 0.44 and was matched with an actual usage of 0.48. OpenMS®,
in conjunction with algorithms designed in house, have allowed for a significantly
quicker selection of two orthogonal columns, which have been optimised for a 2D-
HPLC separation of crude extractions of plant material.
Principal components analysis (PCA) loading plots were used for the first
time to elucidate key differences between two-dimensional high performance liquid
chromatography (2D-HPLC) fingerprint data from samples collected from stages
along the Papaver somniferum industrial process chemistry workflow. Data
reduction was completed using a 2D-HPLC peak picking approach as a precursor
to chemometric analysis. Comparisons of the final stages of product extraction as
an example, PCA analysis of characteristic 2D-HPLC fingerprints accounted for
84.9% of variation between the two sample sets measured in triplicate, with 64.7%
explained by PC1. Loadings plots of PC1 on each sample set identified where the
significant changes were occurring and normalised bubble plots provided an
indication of the relative importance of each of these changes. These findings
highlight 2D-HPLC with appropriate chemometric analysis as a useful tool for the
exploration of bioactive molecules within biomass.
Two-dimensional HPLC and PCA was utilised for the first time here to
develop a map of the morphine and thebaine/oripavine processing streams. In total
68 samples were analysed, including the main processing line, waste streams and
crude impurity standards. The major changes between crucial steps in the
processing were investigated with the biggest changes identified by mass.
Impurities of interest were also investigated to find if they are a consequence of the
extraction process or originate in the natural plant product. These findings will
assist the company in quality control and understanding of the significant impurity
changes throughout the process.
XI
Measures to control the expanding use of synthetic cannabinoids demand
new analytical methodology to identify and determine compounds within this
rapidly evolving class. Seven synthetic cannabinoids (AM-2201, UR-144, XLR-
11, A796,260, 5F-AKB-48, PB-22 and 5F-PB-22) were identified to be present in
eleven herbal products sold in Victoria, Australia, prior to their ban in 2014, using
a combination of GC-MS, HPLC, ESI-MS, and NMR. An additional cannabinoid
in a twelfth product was not detected but not identified. In aid of this work nine
synthetic cannabinoids standards (AM-2201, AB-001, 5F-AB-001, JWH-018,
JWH-073, 5F-AKB-48, 5F-AB-PINACA, 5F-PB-22 and PB-22) were synthesized
or purchased. Furthermore, chemiluminescence detection of synthetic cannabinoids
was investigated using three commonly used reagents: acidic potassium
permanganate, colloidal manganese(IV), and tris(2,2′‐bipyridine)ruthenium(III).
Using the permanganate reagent, no chemiluminescence signal was obtained, but
the manganese(IV) and tris(2,2′‐bipyridine)ruthenium(III) reagents gave
analytically useful responses with all synthetic cannabinoids under investigation
except 5F-AKB-48 and 5F-AB-PINACA. Calibration curves for PB-22, 5F-PB-22,
AM-2201 and 5F-AKB-48, prepared using HPLC with UV absorbance and/or
chemiluminescence detection, were used to determine the total cannabinoid content
extracted with methanol (1 mL) from six of the herbal products (10 mg), which
ranged from 0.082 to 0.77 mg. Limits of detection were determined for all standards
with both UV absorbance and/or chemiluminescence detection, which ranged from
1 × 10-7 M to 9 × 10-4 M. A solid phase extraction study was conducted to further
isolate the synthetic cannabinoid compounds. Six cartridges were trialled and all
were found to clean up the sample, however a C18 silica cartridge proved superior.
Solid-state 13C and 19F NMR spectroscopy has been used for the first time
to detect forensically relevant synthetic cannabinoids on the surface of herbal
substrates. The 13C spectra showed narrow peak widths suggesting that the
synthetic additives are deposited in an ordered, potentially crystalline, phase.
Importantly, the use of solid-state NMR, particularly 19F, offers a relatively fast,
non-destructive, selective detection protocol for synthetic cannabinoids on complex
samples and has potential for other applications such as the direct detection of
pesticides on plants.
XII
ACKNOWLEDGEMENTS
I would like to acknowledge a number of people for all of the guidance and
support they have provided throughout my candidature.
Firstly, to my primary supervisor Dr. Xavier Conlan I would not have done
a PhD if you hadn't offered me this great opportunity and interesting project. I can't
thank you enough for all your support, encouragement and help throughout the
years. We have had some great laughs and hopefully there will be many more
conferences to come!
To my associate supervisor, Alfred Deakin Professor Neil Barnett I am
grateful for the words of wisdom, encouragement and support. Thank you for your
part in convincing me to do a PhD. I will miss the Lakehouse lunches.
To my other associate supervisor Dr. Paul Stevenson, thank you for your
support and help with driving the data analysis portion of my thesis. The data would
not have made sense without your help and it is greatly appreciated. The last few
months without you being down the hall were definitely tough!
To Professor Paul Francis, thank you for all your help with
chemiluminescence and ECL. It has been a great learning experience and I really
appreciate your significant input into the development of that work.
To Sun Pharmaceuticals (formerly GSK) thank you for the opportunity to
participate in such an interesting project. To my external co-supervisor Dr. Stuart
Purcell your wealth of knowledge has been very helpful and you time much
appreciated. Thank you also to Fiona Fry and Tim Bowser for your guidance and
support throughout my candidature.
Thank you to Dr. Zoe Smith for all your assistance over the years. All your
help with the weird and unique instrument problems and chemiluminescence was
much appreciated. Your words of wisdom, guidance and support when needed have
been incredibly valuable.
Thank you to Dr. Jim Pearson for your part in the work presented in
Chapters 5 and 6. I appreciate the time you have taken and input you have had into
XIII
the project. To Dr. Luke O'Dell, Haijin Zhu and Dr. Fred Pfeffer thank you for all
you help with the solid state NMR. The opportunity to use this technique and the
outcomes achieved have been an enjoyable experience.
To Dr. Luke Andrighetto you have been a great desk buddy, the laughs
definitely made the years easier. To Luke and the soon to be Drs Ashton Theakstone
and Natalie Gasz, lunch was always a highlight of the day. Thanks for your
friendship and listening to me vent when nothing was working. We have had some
great times and adventures that I'll never forget!
To the Deakin analytical group past and present, thank you for being fun
and easy to work with and helping shape the memories from lab days to
conferences, it has been a great ride!
A heartfelt thank you to my family especially Mum, Dad, Chrystal, Nan and
Pop for always believing in me and encouraging me to follow my dreams. I could
not have made it this far without your unconditional love and support.
And last but definitely not least my partner in crime, Kris. Thank you for
always being there to support me and listen to me vent. Your love, hugs and ability
to make me forget about bad days have been priceless and I could not have done it
without you!
XIV
PUBLICATIONS
Listed below are references to publications that have resulted from the work presented in this thesis:
Niki K. Burns*, Luke M. Andrighetto*, Xavier A. Conlan, Stuart D. Purcell, Neil W. Barnett, Jacquie Denning, Paul S. Francis, and Paul G. Stevenson, Blind column selection protocol for two-dimensional high performance liquid chromatography, Talanta, 154, 85-91 (2016) DOI: 10.1016/j.talanta.2016.03.056, (*Co-first authors)
This paper is a result of the work presented in Chapter two of this thesis.
Paul G. Stevenson*, Niki K. Burns*, Stuart D. Purcell, Paul S. Francis, Neil W. Barnett, Fiona Fry, Xavier A. Conlan, Application of 2D-HPLC coupled with principal component analysis to study an industrial opiate processing stream, Talanta, 166, 119-125 (2017) DOI: 10.1016/j.talanta.2017.01.044, (*Co-first authors)
This paper is a result of the work presented in Chapter three of this thesis.
Niki K. Burns, Trent D. Ashton, James R. Pearson, I. Leona. Fox, Frederick M. Pfeffer, Paul S. Francis, Zoe M. Smith, Neil W. Barnett, Lifen Chen, Jonathan M. White and Xavier A. Conlan, Extraction, identification and detection of synthetic cannabinoids found pre-ban in herbal products in Victoria, Australia, Forensic Chemistry, Submitted: 28/02/2017
This paper is a result of work presented in Chapter five of this thesis.
Niki K. Burns, Haijin Zhu, Luke A. O’Dell, James R. Pearson, Fred M. Pfeffer and Xavier A. Conlan, Identification of synthetic cannabinoids on herbal substrates by solid-state NMR, The Analyst, Submitted: 2017
This paper is a result of the work presented in Chapter six of this thesis.
Other publications not presented in this thesis include:
Luke M. Andrighetto*, Niki K. Burns*, Paul G. Stevenson, James R. Pearson, Luke C. Henderson, Christopher J. Bowen and Xavier A. Conlan, In-silico optimisation of two-dimensional HPLC for the determination of Australian methamphetamine seizure samples, Forensic Science International. 266, 511-516 (2016) DOI: 10.1016/j.forsciint.2016.07.016, (*Co-first authors)
XV
ABBREVIATIONS
2D-HPLC Two-Dimensional High Performance Liquid
Chromatography
µm Micrometre
ACN Acetonitrile
amu Atomic mass unit
API Active Pharmaceutical Ingredient
Au Absorbance units
CE Capillary Electrophoresis
CHS Cannabinoid Hyperemesis Syndrome
CP Cross Polarisation
CPS Concentrate of Poppy Straw
CV Cyclic Voltammetry
DAD Diode Array Detector
DMF Dimethylformamide
DMSO Dimethylsulfoxide
DNP Dynamic Nuclear Polarisation
ECL Electrochemiluminescence
ESI-MS Electrospray Ionisation-Mass Spectrometry
FC Ferrocene
FIA Flow Injection Analysis
GC Gas Chromatography
GC/MS Gas Chromatography Mass Spectrometry
GSK GlaxoSmithKline
XVI
HPLC High Performance Liquid Chromatography
LC Liquid Chromatography
LC/MS Liquid Chromatography Mass Spectrometry
LOD Limit of Detection
M Molar/ moles per litre
MAS Magic Angle Spinning
Mn(IV) Manganese(IV)
MS Mass Spectrometry
MSD Mass Selective Detector
MW Molecular Weight
m/z Mass to charge ratio
NMR Nuclear Magnetic Resonance spectrometry
PCA Principle Components Analysis
PMT Photo Multiplier Tube
ppm Parts per million
psi Pounds per square inch
QCA Quality critical attributes
QCP Quality critical parameters
RT Retention Time
[Ru(bipy)3]3+ Tris(2,2-bipyridine)ruthenium(III))
SDS Safety Data Sheets
SPE Solid Phase Extraction
SPIA Sun Pharmaceutical Industries Australia
THC Tetrahydrocannabinol
1
CHAPTER 1:
INTRODUCTION
2
Complex samples in context
Complex samples have long been an issue in analytical chemistry and are
increasingly common due to improvements in instrument sensitivity [1]. Highly
sensitive methods mean that increasingly low amounts of analyte can be determined
leading to identification of previously undetermined contaminates. The isolation of
drugs from complex samples is a priority in both forensic and pharmaceutical
disciplines, and can be particularly difficult when samples are obtained from natural
plant sources [2, 3]. Natural product extractions for purposes such as
pharmaceuticals and food may still contain numerous components, requiring
multiple steps to purify the target analytes [3]. With the improvement in impurity
profiling, demands are placed on developing protocols to remove these compounds.
Also known for sample complexity, testing in the forensic discipline ranges from
seized crude drugs [4] to biological samples [5] where the target compound is only
a small portion of the sample. Seized drugs often contain excipients, precursors and
bulking agents in addition to the main drug which demands the analytical
techniques be refined to allow efficient analysis of these and other complex
samples.
This thesis examines two types of drug sample matrices in detail: opiates
from poppy straw obtained from Sun Pharmaceuticals and synthetic cannabinoids
sprayed onto herbal substrates, in conjunction with Victoria Police Forensic
Services Department. Sun Pharmaceuticals, formerly known as GlaxoSmithKline
are a pharmaceutical company that employ a complicated multistep extraction
process to isolate opiates of commercial interest. Synthetic cannabinoids have
relatively recently been banned in Victoria [6], prompting Victoria police to
investigate these complex samples in order to develop robust separation and
identification protocols.
Opiates
The opium poppy, thought to have originated in Asia Minor, is one of the
oldest plants recorded for medical use [7] with evidence suggesting that it could
have been used as a pain reliever or as a drug of abuse for thousands of years. The
opium poppy has been referenced in Sumerian clay tablets dating back to 3000BC
3
and in Greek mythology in 850BC [7]. Archaeological evidence has also
discovered their presence from the Minoan civilisation, dating back to the 14th
century BC [8].
Alkaloids can be extracted from many different poppy species including
Papaver Bracteatum, Papaver Rhoeas and Papaver Setigerum, however, Papaver
Somniferum is the species cultivated for opium production as it produces the most
viable amount of opiate products [9]. The opium poppy plant produces natural
opiate alkaloids, such as morphine and thebaine and in 1806 Sertürner published
his paper on isolating the first alkaloid (morphine) from the opium poppy [10].
After morphine was isolated, other opiates (noscapine, papaverine and thebaine)
were identified with codeine being isolated in 1833 [10]. Morphine is the highest
concentrated opiate in the species Papaver somniferum making this an industrially
significant plant. Metabolically engineered strains of the plant have been developed
[11] allowing a crop of poppy plants to selectively produce more of specific opiates.
This has been achieved with selection [12], cross pollination [13], mutation
breeding [14, 15] and more recently metabolic engineering [16]. Production of
thebaine and oripavine are among the opiates that have expanded from poppy
hybrids [14].
The industrial scale extraction of these alkaloids from the plant is
commonly done one of two ways. The traditional way involves slicing the capsule
of the opium poppy and collecting the gum (latex) that oozes out to seal the wound
and is typically collected 12 - 24 hours later. The latex is then air dried over
approximately 10 days before the crude opium is stirred every half an hour over a
1-3 week period in the sun [7], before following a similar method to the following
technique for isolation of morphine. This method is still used in northern Indian
regions however it is outdated and inefficient and as such is not used in modern
processing [7]. The most common extraction method is the use of poppy straw,
where 83% of the morphine and 91% of thebaine manufactured worldwide utilise
this approach [17]. The poppy straw contains the upper stalk of the plant including
the crushed capsule that has been dried and made into straw pellets. These straw
pellets are subjected to a water and organic solvent extraction process to isolate the
pharmaceutically beneficial opiate compounds in bulk. This method is utilised by
Sun Pharmaceutical Industries Australia for their morphine and thebaine crops,
4
with other target opiates being extracted and further processed for purification. The
four common alkaloids morphine, codeine, oripavine and thebaine are of interest
in this thesis and are illustrated in Figure 1.1.
a) b)
c) d)
Figure 1.1: Structures of the four major opiates a) Morphine, b) Codeine, c) Oripavine and d) Thebaine.
Morphine, the main opiate found in the opium poppy (Papaver
Somniferum), is used as a pain reliever or analgesic for a variety of different types
of pain and is commonly used in a hospital setting to bring effective relief from
severe pain [18]. Common side effects of morphine include sedation, euphoria,
constipation and a lowered blood pressure [7] however the euphoric and addictive
effect of this drug has also led to the abuse of this medicine and consequently it is
now a controlled substance along with other opiates [19]. Morphine is commonly
5
harvested as a precursor to the illicit drug heroin (diacetylmorphine), a compound
that is also legally manufactured in the UK for medical purposes to treat severe pain
in particular for terminally ill patients, during caesarean sections and other major
surgeries [20].
Codeine is a commonly prescribed analgesic and is found in cold and flu
medication or strong pain relief tablets and is suited to relieve mild to moderately
strong pain [7]. Oripavine is extracted from an opium poppy that has been
genetically modified to produce the opiate in a larger quantity and is then used to
produce other pharmaceutical drugs such as buprenorphine and naltrexone [21].
Oripavine is unable to be used as an analgesic despite having an effect comparable
to morphine, due to its toxicity [22]. The difference in toxicity is noted in safety
data sheets (SDS), where the LD50 of morphine is 250 mg/kg (rat) compared to
oripavine with an LD50 of 26.1 mg/kg (intraperitoneal, rat). Likewise, thebaine is
unable to be utilised due to its high toxicity (LD50 of 114 mg/kg, rat, oral) that
causes strychnine-like convulsions [23, 24]. Hence it is extracted from the opium
poppy as a precursor to other pharmaceutically important drugs such as oxycodone,
oxymorphone, naltrexone, nalbuphine, buprenorphine and etorphine [25].
Opium poppies are grown on a large scale in countries including Australia,
India, and Turkey for the pharmaceutical industry. The multinational company Sun
Pharmaceutical Industries are a medicinal based company responsible for the
production of morphine and other opiates used in pharmaceuticals. Heritage
companies to Sun Pharmaceutical Industries Australia (SPIA) have been operating
in Australia since 1886. In 1964, Glaxo began growing opium poppies in Tasmania
for the extraction of morphine and other related opioid medications [26]. Since then
the company has grown significantly, contributing to the global supply of opiate
pharmaceuticals and in 1995 the company merged to form Glaxo Wellcome and
more recently GlaxoSmithKline (GSK) (2000-2015). In 2012 GSK invested
AUD$54 million to research and development [26] with trials in Victoria to
determine the viability of poppy production on mainland Australia. In 2015 Sun
Pharmaceuticals purchased the opiate section from GSK and since then has
continued with the cultivation and processing of important opiate alkaloids. The
core opiates isolated by Sun Pharmaceuticals are morphine, codeine, oripavine and
thebaine (Figure 1.1) for use in various medicinal products or as a precursor to
6
many important semi-synthetic opiates. As these bulk opiates are being utilised in
consumable products it is important to monitor the process for impurities. A
mapping technique for the extraction process will also help to inform the company
of what is occurring at critical points in the stream thus facilitating quality control.
In general mapping techniques allow visualisation of changes by using the data
collected to identify differences between data sets or in this case, each step of
manufacturing processes.
Synthetic cannabinoids
Synthetic cannabinoids, originally based on naturally occurring
cannabinoids, were developed as therapeutic agents and possible alternatives to
THC [27, 28]. However, these failed to diminish the side effects already commonly
observed from the natural cannabis plant. The first instances of these molecules
were published in academic journals after development and investigation by
pharmaceutical industries and researchers in the 1960’s [29, 30]. John W Huffman,
the inventor of several commonly seized synthetic cannabinoids is also the name
sake of the compounds starting with JWH, such as JWH-018 [31, 32]. Other
compounds are either named after their inventor, their base structure or the class in
which they belong. Examples of this include CP-47,497 which whilst invented by
Pzifer, is named after its cyclohexylphenol base structure, as are other synthetic
cannabinoids described with the CP prefix [29].
Since the early 2000s in Australia [27, 33] a large number of products
marketed as "legal highs" were readily available in herbal shops and adult stores.
However, in 2008 these herbal substrates were tested and found to contain
previously reported synthetic cannabinoid compounds as the active ingredient [29,
34]. These products consisted of herbal substrates that had been laced with one or
more synthetic cannabinoid compounds and were then smoked in a similar manner
to cannabis. These chemicals are commonly white powders or oils that are
dissolved into acetone or a similar organic solvent and sprayed onto the herbal
substrates [35, 36]. These substrates are commonly made from the leaves, bark,
flowers and roots of plants such as damiana (Turnera Diffusa), marshmallow leaf
(Althaea officinalis) and mullein leaf (Verbascum Thapsus) [37, 38].
7
To date hundreds of different synthetic cannabinoid structures have been
identified making the detection and identification of these compounds a difficult
task, and law enforcement agencies are in need of improved protocols [27, 39]. A
common way to avoid the legislation against the compounds was to make analogues
or change the structures in such a way that they were no longer illegal. As is the
case with many of the compounds, an example being UR-144 and XLR-11 where
the latter has only the addition of a fluorine on the end of the alkyl chain (Figure
1.2). This is a common analogue along with changes to the aromatic rings or the
addition of extra carbons on the alkyl chain. This has made the detection of these
difficult as not all structures have the same properties, prompting a push to find
alternative/new analytical methods that can selectively detect these compounds.
Figure 1.2: Structures of UR-144 and its fluoro pentyl analogue XLR-11.
To further complicate the chemical fingerprint of these products, fragrances
and flavourings such as menthol, strawberry, mango and pineapple are added to
some brands to give a higher appeal to the purchaser. These products, while
marketed as legal highs, were sold under the radar as potpourri and incense and
their devastating impact on society is illustrated in Figure 1.3 [40-42]. Most
products contain labels stating the herbs were ‘not for human consumption’ and
‘for aromatherapy use only’ with some blends including warnings of the potential
to get a high, as is illustrated in Figure 1.4 with the brand Cloud 9.
8
Figure 1.3: A snapshot of news headlines relating to the emergence of synthetic cannabinoids and the effects caused by these drugs.
9
Figure 1.4: Warning label on the brand Cloud 9 stating “WARNING: This product is intended to be used as incense, for Aromatherapy use only. The deliberate smoking of
this product may result in you getting high and experiencing hallucinations or delusions.”
Synthetic cannabinoids have a similar effect in the body as
tetrahydrocannabinol (THC) as these compounds interact with the same chemical
receptors in the brain, Cannabinoid 1 (CB1) and Cannabinoid 2 (CB2) [28, 43-45].
A series of emergency department case studies have found users suffer from
anxiety, tachycardia, tachypnoea, hypertension, hallucinations or psychosis and
seizures [46]. Hopkins and Gilchrist [47] reported a link between these compounds
and cannabinoid hyperemesis syndrome (CHS). A common problem found in
heavy marijuana smokers, CHS is believed to be caused by an over stimulation of
the CB 1 receptor [47]. As these products gained popularity, more cases of adverse
side effects were reported with numerous deaths and over doses recorded as
illustrated in Figure 1.3 by a snapshot of Australian news headlines [48]. Fatal
intoxication associated with these synthetic drugs has occurred due to organ failure,
where in particular the kidneys and heart are irreparably damaged [49]. A
newspaper article in the Geelong Advertiser in December 2013 reported on three
people in the Geelong Hospital intensive care unit suffering with organ failure after
using synthetic cannabinoids and numerous others that had reported to emergency
in the preceding weeks [41].
As synthetic drugs such as cannabinoids and cathinones are a new craze,
statistics are yet to reflect the full extent that these products are having on society.
In 2013 the National Drug Strategy Household Survey (NDSHS) [50] included new
psychoactive drugs, including synthetic cannabinoids, into the survey for the first
time. This survey found that 1.3% of Australians aged 14 and over had used
10
synthetic cannabis in the past 12 months, with the most common age group being
the 14-19 year old range. The Australian data from the Global Drug Survey found
that in 2014 synthetic cannabis was the 20th most used drug with 4.1% of
respondents having used the drug in the last 12 months [51]. Even more alarming
is the findings of the 2016 Global Drug Survey [51] where it was found that for the
fourth year in a row synthetic cannabinoids have topped the list of drugs ‘leaving
you most likely to need emergency medical treatment’. The report found that 1 in
30 people who had used in the last year required emergency medical treatment.
This is a significant number for a drug that has only been around for 13 years and
highlights the importance of controlling such dangerous substances.
In May 2014, a ban was put in place in Victoria, Australia to prevent the
purchase, sale and possession of 13 classes of synthetic cannabinoids that included
any analogues of these compounds that fall into one of the specified classes under
an amendment to the Drugs, Poisons and Controlled Substances Act 1981 [6]. This
new legislation took into account the previous prohibition of 8 synthetic
cannabinoids in 2011 [52], and introduced the analogue approach which targeted
small changes to the molecules that had previously led to the evasion of the initial
ban. As these compounds are still an issue in Victoria further action is in progress
with the Drugs, Poisons and Controlled Substances Miscellaneous Amendment Bill
2017 that introduces a blanket ban of all ‘psychoactive substances’ as was detailed
in a press release in March 2017 [53], however the bill is still progressing through
parliament. Due to these recent bans methodology for the detection of these
compounds needs to be developed.
Separation and Detection methods
Separation and detection methods for complex samples such as natural plant
products and synthetic mixtures typically involve techniques including
chromatography and spectroscopy. Synthetic cannabinoids and opiates have been
detected with methods such as Mass Spectrometry (MS) [54], Chemiluminescence
[55] and Nuclear Magnetic Resonance (NMR) spectroscopy [3] post separation
with analytical systems such as High Performance Liquid Chromatography (HPLC)
11
[17, 37], Gas Chromatography (GC) [56] and Capillary Electrophoresis (CE) [57].
A summary of each of these important analytical instruments, including detail about
HPLC, is outlined below with a targeted summary given in the introduction section
of relevant results chapters.
High Performance Liquid Chromatography (HPLC)
High Performance liquid chromatography (HPLC) is a major tool in
analytical chemistry and contemporary science for the separation of a vast array of
compounds [58]. The applications of this already popular technique has steadily
increased in recent times as a review by Gika et al. in 2015 highlighted, by the
increase in the number of papers published utilising HPLC (Figure 1.5) [58].
Reverse phase is the most common HPLC system where a polar mobile phase and
a non-polar stationary phase are utilised to separate chemical components [59]. This
chromatography technique separates analytes based on their hydrophobicity due in
the case of the C18 stationary phase for example to the methylene selectivity of the
reverse phase column, where polar compounds are eluted first followed by non-
polar compounds [60].
Figure 1.5: An illustration by Gika et al. describing the number of papers published utilising high performance liquid chromatography [58] (where liquid chromatography and one of the four traces in the graph were searched in the Scopus database in 2014).
12
Reverse-phase HPLC has been used previously for the separation of
synthetic cannabinoids with the majority of published studies utilising solvent
gradients of either acetonitrile or methanol in water, with modifiers such as formic
acid [54]. Other solvent systems employed for the determination of synthetic
cannabinoids include a gradient of 1:1 methanol: acetonitrile in water [38], and an
isocratic aqueous sodium acetate buffer [61].
Reverse phase HPLC is also the method of choice to analyse opiates
including morphine, thebaine and buprenorphine [62]. Acevska et al. developed
and validated a reverse phase HPLC method for the resolved separation of six
targeted opiates [17]. Methods developed for Sun Pharmaceutical Industries
Australia (formerly GSK) utilise reverse phase HPLC for the analysis of the
alkaloid processing streams [63].
Two-Dimensional High Performance Liquid Chromatography (2D-HPLC)
In some cases the separation power of a one dimensional system is not
enough to resolve all components of a sample [64]. In this case two-dimensional
High Performance Liquid Chromatography (2D-HPLC) has the ability to achieve a
greater separation power by increasing the peak capacity [65]. Two-dimensional
High Performance Liquid Chromatography is a technique that utilises two
orthogonal (different retention mechanism) columns to achieve a greater separation
potential [66]. This technique can be performed in either offline or online methods.
Offline 2D HPLC requires the first dimension to be run and the fractions collected
into vials; subsequently the samples in these vials are then manually reinjected into
the second dimension [67]. The online method can be done comprehensively or via
heart cutting, where both utilise a switching valve and sample loops to store
fractions before transferring these fractions into the second dimension as illustrated
in Figure 1.6 (a diagram of the system used by Stevenson et al. at Deakin
University) [68]. In comprehensive 2D-HPLC the whole first dimension is
transferred into the second dimension whereas in heart cutting only selected peaks
are forwarded to the second dimension [64]. Online 2D-HPLC is a quicker and
more efficient method than offline, however it requires the second dimension to
have a fast flow rate as, in order to avoid peak overlap, the second separation must
be complete before the first dimension re-fills the sample loop [67].
13
Figure 1.6: Two-position, eight-port switching valve configuration utilised in 2D-HPLC separations as described by Stevenson et al. [68]
Other applied separation techniques
Other separation techniques that have been applied to the substrates
analysed here include gas chromatography and capillary electrophoresis, however
they have not been used in this thesis due to the limitations outlined in this section.
Gas chromatography (GC) is a separation technique commonly used for the
analysis of a wide variety of volatile compounds. The use of GC is widespread in
chemistry with applications in forensics [5, 69, 70], pharmaceuticals [71, 72] and
quality assurance in chemical industries [73], where it is often coupled to mass
spectrometry (GC-MS). It is essential for GC that the samples are volatile and that
the compounds to be analysed do not degrade with high temperature. This is an
issue with some synthetic cannabinoids such as XLR-11 which contain a
cyclopropyl ring as the ring opens to form a propenyl rearrangement product of the
original compound when heat is applied [74].
Capillary electrophoresis (CE) is an analytical technique utilised for
applications such as pharmaceuticals [57, 75] and forensics [76] where charged
analytes are separated based on their differing electrophoretic mobility’s. Hindson
et al. has previously used this technique for the determination of opiate alkaloids in
14
process liquors [77]. However a limitation of this technique is the lack of sensitivity
or stability often required for forensic applications and as such it is not well
established in these laboratories [76].
Chemiluminescence
Chemiluminescence is the production of light from a chemical reaction
without the generation of heat [78]. In an analytical chemistry setting the light is
commonly detected with a photodetector as over a finite range the intensity of the
light is proportional to the concentration of reactants involved [79]. It is important
to note that the amount of light produced is not necessarily about how well the
reagent reacts with an analyte but is due to the production and stability of the
excited state. This is a system that is easily coupled to HPLC [80, 81] or Flow
Injection Analysis (FIA) [82, 83] and may be used in conjunction with a UV-
absorbance detector [63, 84].
There are a range of different chemiluminescence reagents that can be
utilised for detection and they are typically prioritised based on the chemical
structure of the target analytes, as each reagent is generally known to elicit light in
the presence of particular functional groups. [79]. The most common
chemiluminescence reagents utilised at Deakin University are tris(2,2‐
bipyridine)ruthenium(III) ([Ru(bipy)3]3+), colloidal manganese(IV), acidic
potassium permanganate and luminol of which the preceding three have been
utilised in this thesis (Figure 1.7). The use of chemiluminescence is ideally suited
where selectivity is required. For example the opiates morphine and codeine are
similar in structure, however they react differently with the aforementioned
reagents. As described by Francis et al., codeine produces an intense emission with
[Ru(bipy)3]3+ where a relative intensity of 100 was obtained, but affords minimal
light with acidic potassium permanganate with a relative intensity of only 2
observed [55]. Morphine however is the opposite with the relative intensity with
acidic potassium permanganate observed at 100 as opposed to less than 0.01 with
the [Ru(bipy)3]3+ reagent [55]. This is proposed to be due to the lack of a phenol in
codeine that morphine contains, an attribute that elicits light with acidic potassium
permanganate but quenches [Ru(bipy)3]3+ [79]. However, Lenehan et al. overcame
15
this selectivity by utilising a mix of these two reagents to detect both morphine and
codeine in the same system [85].
a) b) c)
Figure 1.7: Structure of a) potassium permanganate, b) [Ru(bipy)3]3+ [55] and c) manganese dioxide (Mn(IV)).
A review on chemiluminescence by Adcock et al. highlights the sensitivity
of the technique with detection limits as low as 1 × 10-11 M reported. Table 1.1
highlights the use of the following three reagents: [Ru(bipy)3]3+, acidic potassium
permanganate and colloidal magnesium(IV) for the detection of a range of dugs.
Table 1.1: Detection limits of a range of opiate based molecules gained
with three chemiluminescence reagents.
Analyte Reagent Instrumentation LOD
Codeine Manganese(IV) FIA 1.0 × 10-8 M [86]
Thebaine Manganese(IV) FIA 5.0 × 10-9 M [86]
Norfloxacin Manganese(IV) FIA 3.0 × 10-8 M [87]
Cannabidiol Acidic potassium
Permanganate
HPLC 1.0 × 10-6 M [88]
Oripavine Acidic potassium
Permanganate
HPLC 1.0 × 10-9 M [89]
Morphine Acidic potassium
Permanganate
HPLC 7.5 × 10-10 M [89]
MDMA [Ru(bipy)3]3+ HPLC 4.8 × 10-7 M [90]
Codeine [Ru(bipy)3]3+ HPLC 5.0 × 10-10 M [63]
Thebaine [Ru(bipy)3]3+ HPLC 1.0 × 10-9 M [63]
16
[Ru(bipy)3]3+ is particularly useful for the detection of alkaloids,
biomolecules, pharmaceuticals and controlled drugs containing secondary or
tertiary amines [90-93]. The [Ru(bipy)3]2+ is first oxidised to [Ru(bipy)3]3+ and then
reduced by an analyte to produce light and [Ru(bipy)3]2+ 79 . Recently, a highly
stable form of this reagent was developed by McDermott et al. [91], where the
complex is prepared as a perchlorate salt which allows the reagent to remain in the
3+ state for a longer period of time. This reagent has been found to react strongly
with tertiary amines [78].
Acidic potassium permanganate was first reportedly used in 1917 to study
the oxidation of pyrogallol [94] and was also the first reagent used to determine
morphine in 1986 [95]. The acidic potassium permanganate also known as
manganese(VII) is proposed to form an excited state manganese(II) species via a
multi-step reduction with certain analytes, in particular those that contain phenols
[96]. This reagent has a wide range of analytical applications and has been used
extensively in pharmaceutical analysis, in particular opiates such as morphine, and
for illicit drugs, biomolecules, pollutants, antioxidants and pesticides [94, 96].
The colloidal manganese(IV) reagent was first reported in 2001 and was
found to produce light with many organic and inorganic compounds when used in
conjunction with an enhancer such as formaldehyde or ethanol [97]. This reagent
has a similar mechanism to acidic potassium permanganate where manganese(IV)
is reduced to manganese(II). This reagent exhibits a broader selectivity than that of
acidic potassium permanganate and reacts with amino acids, alkaloids, indoles,
plant-based extracts and analytes of forensic interest [86, 97-101]. This broader
selectivity is demonstrated with opiate alkaloids where it was found to react with
both morphine and codeine as opposed to just one as is noted with the previously
discussed reagents.
Electrochemiluminescence (ECL)
Electrochemiluminescence (ECL) is a technique that involves applying a
potential to an electrode in the presence of a reagent and co-reactant to induce an
electron-transfer reaction to form light emitting excited states[102]. It was first
reported for use with Ru(bipy)22+in 1972, where the sample was in acetonitrile with
the electrolyte tetrabutylammonium tetrafluoroborate [103]. It is commonly used
17
in biological and clinical testing [104], food and water assessment [105] and drug
detection [106]. Microfluidic analytical devices have also used ECL as a detection
method [104]. This technique is commonly used due to its high sensitivity and
selectivity that it affords an anlalyst [102, 104].
Mass Spectrometry (MS)
Mass spectrometry (MS) is a technique used for the identification of
compounds by determining the molecular mass and the mass of fragment ions
[107]. There are three main processes within the mass spectrometer; 1) the sample
is first ionised and then 2) passed through an electric field, ionic field, magnetic
sector or drift region where the ions are separated before 3) detection [108]. Mass
Spectrometry is a technique highly utilised in analytical chemistry for the detection
and identification of compounds of interest. It is heavily utilised in the
pharmaceutical industry for the determination of impurities and degradation
products and is often coupled with GC (gas chromatography), LC (liquid
chromatography) or CE (capillary electrophoresis) [71]. It is also used extensively
in forensics for biological samples and the identification of drugs or other
compounds.
Mass spectrometry is used in the forensic analysis of synthetic cannabinoids
and is the main identification technique utilised in conjunction with gas
chromatography [54]. This technique has been instrumental in identifying new
compounds where it is coupled to either LC or GC [109-112] and in biological
testing where it has been used to test for synthetic cannabinoids in substrates such
as blood [113-115], hair [116, 117] and urine [118, 119].
Extensive literature is available on the use of mass spectrometry for opiates
in biological samples [120-122] and in environmental applications to test their
presence in waste water [123-125]. MS has been instrumental in mapping the
opium poppy biosynthesis pathways [126] and for identifying opiate alkaloids after
isolation [127]. Stranska et al. utilised HPLC/MS/MS for the determination of
alkaloids from empty Papaver somniferum capsules [128]. This technique is most
commonly used in conjunction with liquid chromatography, where LC/MS is
commonly used on sites that extract bulk opiates as a quality control measure.
18
Nuclear Magnetic Resonance Spectroscopy (NMR)
Nuclear Magnetic Resonance (NMR) Spectroscopy is a technique often
employed for structural determination of a compound. This method has been widely
employed in both synthetic and analytical chemistry, for structural elucidation or
confirmation of compounds [108]. NMR is most commonly performed in a solution
state using a deuterated solvent, however compounds can also be analysed in solid
state [129]. Solid state NMR is an emerging technique that is often used for the
analysis of bulk samples and to investigate the surface chemistry of catalysts,
zeolites and metal organic frameworks [130-132].
Synthetic cannabinoids have evolved greatly since the discovery of the first
compounds on marketed "legal high" products. In conjunction with mass
spectrometry, these emerging compounds are commonly identified with NMR.
There are a number of papers reporting the use of 13C, 19F and 1H solution state
NMR for the structural elucidation of these new compounds [110, 133-135]. Fowler
et al. used 1H NMR and proton correlation spectroscopy to quantitatively identify
synthetic cannabinoids on herbal blends [136]. Opiate alkaloid compounds have
also previously been determined using techniques that include NMR [137, 138].
Furuya et al. utilised this technique to identify a new alkaloid present in the callous
tissue of Papaver Somniferum [127].
Data Analysis
The data obtained from LC/MS and HPLC can often be considerable and
difficult to handle when dealing with a large number of samples. Due to this, data
handling software is often employed to minimise the time it takes to interpret the
information. There are numerous open source software, such as OpenMS® and
Open chrome, available in addition to the programs supplied by the manufacturers
of the instrumentation, namely Mass hunter (Agilent), Profiling solutions
(Shimadzu) and Xcaliber (Thermo Fisher). Mathematica and other programming
software such as Matlab, allow for better interpretation of complex mass
spectrometry data and 2D- HPLC. With algorithms designed in house these
19
programs have the potential to determine important data from a large dataset and
compose heat maps for 2D separations.
Principal components analysis (PCA) is a common statistical technique to
determine differences between sample sets and identify the variables. This
technique was first reported in the literature by Hotelling in 1933 and has since
been utilised for a vast array of applications. It has previously been exploited for
analysis of pharmaceuticals in rats [139], water quality testing [140], food testing
[141, 142] and wine testing [143, 144].
Project aims
This project aims to develop improved analytical separation and detection
capacity for compounds in complex samples. The basis for this work is to
interrogate complex samples associated with forensic drug detection and
manufacture of licit drugs in order to enable these industries to best deal with
priority chemical challenges.
An efficient method for the selection of orthogonal columns will be
explored using a combination of multidimensional chromatography, mass
spectrometry and principal components analysis. The aim is to provide a foundation
for the development of a separation designed for complex opiate processing stream
samples. An investigation of all stages of the manufacturing process starting at the
final product and working back to the starting material, raw poppy straw (Papaver
somniferum) for the opiate processing streams will be aimed at mapping the
components in samples from key processing steps. The aim of this is to determine
if compounds of interest had originated from the raw product or as a consequence
of the manufacturing process. The work presented here aims to address the mapping
of the complex opiate manufacturing process so that process owners and industrial
chemists can respond more efficiently to changes identified in the process stream.
In the forensic science arena this work aims to exploit novel analytical
chemistry based on chromatography, chemiluminescence detection and solid state
nuclear magnetic resonance spectroscopy to selectively determine the presence of
synthetic cannabinoids on herbal substrates. The project aims to elucidate the
20
chemical composition of synthetic cannabinoids obtained pre-ban in Victoria
Australia in order to understand the current state of play with synthetic
cannabinoids in this region. The aim of this component is to understand the
structure function relationship between a variety of chemiluminescence reagents
and the synthetic cannabinoids in order to work towards at-scene detection
chemistry. This aim is to be achieved by comparison of the real world synthetic
cannabinoids samples and standards purchased or synthesized at Deakin
University. The fact that synthetic cannabinoids are sprayed onto a herbal substrate
enables the potential exploitation of solid state nuclear magnetic resonance
spectroscopy to probe these samples. For the first time solid state NMR will be
utilised to interrogate synthetic cannabinoids and assess it as a methodology that
may be useful for policing services.
Due to the broad nature of the two components of this thesis namely the
forensic and industrial chemistry more detailed introductory material is given at the
start of each experimental chapter. This has been performed to inform the reader of
the key areas of literature relevant to each of these bespoke areas of scientific merit.
21
CHAPTER 2:
BLIND COLUMN SELECTION PROTOCOL FOR TWO‐DIMENSIONAL HIGH PERFORMANCE LIQUID
CHROMATOGRAPHY
22
Chapter Overview
This chapter explores the use of an open source program, OpenMS,
for handling large amounts of mass spectrometry data, to aid the comparison of
chromatography columns in order to select a pair that will provide the maximum
use of 2D-HPLC separation space. A greater separation space is represented by a
large fcoverage indicating that the chromatographic system allows for better resolution
of compounds in complex samples. The separation was developed using a crude
morphine poppy extract to represent a real world complex sample matrix. Solvent
mismatch was investigated in the development of a suitable 2D-HPLC method.
Once developed, this method was applied to a range of process stream samples
(morphine, thebaine and oripavine poppy extracts) with the fcoverage for each
determined in order to fully assess the suitability of the protocol.
23
Introduction
High performance liquid chromatography (HPLC) is a technique commonly
used for separating compounds in complex substrates [145] such as biological
[146], food [147] and environmental [148] samples. The complexity of a sample
often exceeds the separation potential of traditional HPLC, thus two-dimensional
HPLC (2D-HPLC) is often performed through separating compounds with the aid
of more than one chromatography column [149]. The ability to separate complex
mixtures is greatly increased by using 2D-HPLC as the theoretical peak capacity is
the product of each individual dimension [149]. However, to achieve the best
performance, it is vital that the separation chemistries offered by the two stationary
phases selected are as different (orthogonal) as possible. Consequently, the process
of achieving this is often onerous, requiring a trial-and-error approach or
laboriously injecting a series of representative standards [66, 150].
Many disciplines have utilised 2D-HPLC including pharmaceutical [151,
152], forensic [153], environmental [150, 154] and food sciences [68]. As noted
above, the separation potential of a 2D-HPLC system is determined by the peak
capacity that is used to describe the chromatographic resolving power [155].
Resolution or resolving power of a chromatographic system is defined as the ability
for peaks to be both Gaussian and base line separated. The peak capacity is
proportional to the fractional coverage defined by Watson et al. [156] according to
Equation 1:
∑ (Equation 1)
where fcoverage is the proportional surface coverage that ranges from 0 to 1, ∑bins is
the number of bins (the segmented components of the analogue data signal) that
contain peaks, and P2 is the total number of available bins. The proportional surface
coverage is directly related to the two-dimensional peak capacity, Equation 2,
which illustrates that the full advantage of a 2D-HPLC separation will only be
realised by maximising the fcoverage term [155]:
24
,, ,
(Equation 2)
where, nc,2D is the practical 2D-HPLC peak capacity, nc,i is the peak capacity in
separation dimension i, fcoverage is the proportional amount of separation space
utilisation, and is the under-sampling correction factor, which is directly
proportional to the modulation frequency per peak from the first dimension to the
second, as described by Li and co-workers [157]. This is achieved through a given
number of bins used to divide the separation space, Σbins, that is equal to the
number of peaks; the area of all normalised bins containing peaks is then totalled
giving P2 [158]. Using this method, the fcoverage is calculated as a number between
0 and 1 with the value of 1 indicating full use of the separation space. Gilar et al.
[158] illustrated the use of separation space as shown in Figure 2.1, where it is
important to achieve a separation that is not in a diagonal line and utilises the space
available in the square otherwise low orthogonality is achieved.
25
Figure 2.1: Representation of separation space utility by Gilar et al. [158] where (A) is a non-orthogonal separation representing 0% orthogonality, (B) is a hypothetical, full use
of separation space and (C) is an ideal system representing 100% orthogonality.
26
The separation space is maximised using orthogonal columns; however,
finding two columns that fit this requirement can be a lengthy and difficult task
[66]. Previously, analysing the relative retention profile of each HPLC column
against each other was the only way to identify the best separation. Murahashi and
co-workers [150] selected columns based on differing elution orders by injecting
five standards into eight columns. This was further explored by Bassanese et al.
[66], but it was found that matching peaks by their area with HPLC proved
problematic between different columns when using the sample itself as the
optimisation matrix. Moreover, it was not practical when performing single
injections of a comprehensive series of representative standards. Other methods
include selecting column stationary phases on the knowledge that they are unlikely
to produce similar separations based on the interactions of the compounds being
analysed [68]. These techniques were found to be incredibly time consuming and
difficult to achieve when the components of the sample being analysed are
unknown [66].
For an unknown mixture, one way to match peaks from different separations
is to compare their molecular weight. Mass spectrometry is a technique used for the
identification of compounds by determining the molecular mass and is particularly
important when a greater understanding of the analytes is required [159]. Mass
spectrometry is superior to UV-absorbance detection when analysing complex
sample extractions as it provides an information rich snapshot of the elution such
as a mass associated with each compound in the mixture. This technique also allows
a fingerprint of the sample where the fragmentation pattern can discriminate
between compounds of the same mass [160] thus providing greater information on
the components in complex mixtures than UV-absorbance detection affords. Mass
spectrometry is often coupled with LC (liquid chromatography), GC (gas
chromatography) or CE (capillary electrophoresis) to enhance knowledge of the
concomitant species of a sample [161]. This is particularly important for unknown
samples where as much information as possible is essential for further analysis.
This technique has been utilised in this selectivity study to provide as much
information about the peaks as possible, which can then be used to compare the
columns via the masses of the peaks in the sample.
27
One of the greatest challenges in dealing with mass spectrometry is
handling the enormous amount of data generated. This problem is further
exaggerated by the creation of two-dimensional separations. Consequently,
OpenMS® (version 1.11.1), a program used to manage and analyse data obtained
by mass spectrometry, will be explored to interrogate the complex set of
chromatographic information. In comparison to other programs that are available
for mass spectrometry data analysis, OpenMS® offers a more flexible platform as
the user is able to combine components of the software to suit their individual
requirements [162]. Using this program, filters can be applied to the mass spectra
to extract important information, such as particular peaks of interest. There are a
large number of pipelines that can be followed to obtain the desired outcome for
the analysis, including a simple pipeline that eliminates noise from the spectra or
increasingly more complex pipelines that incorporate a number of filters and peak
picking tools. Visualisation at stages through the pipeline with the TOPPView
feature is also an important element, allowing the user to open a single spectrum or
a peak map in either two or three dimensions [163].
A major difficulty in the optimisation of 2D-HPLC operating conditions is
the selection of columns with divergent retention mechanisms [66]. The aim of this
work is to introduce a new approach to remove the guesswork in column selection;
the methods developed in this study are not limited to alkaloids but represent a
reliable column selection protocol that can be applied to any complex sample where
the constituent components are unknown. This chapter presents an approach to
optimise column selection in 2D-HPLC with a wide range of applications, and is
illustrated here by the separation of crude opiate extracts. Sixteen columns were
used to separate the components within the concentrate of (morphine) poppy straw
(CPS) using LC/MS. OpenMS® was used to evaluate data from each column and
determine which two columns together will provide the best use of separation
space. The selected columns were then used to develop an appropriate 2D-HPLC
method that was applied to additional opiate samples. This method took into
account solvent mismatch from the first to second dimension and optimised both to
allow for the requirement of a fast flow rate in the latter.
28
Experimental
Chemicals and samples
Water was obtained from a Milli-Q filtration system (Merck Millipore,
Victoria, Australia). HPLC grade acetonitrile (ACN) was purchased from Ajax
Finechem Pty. Ltd. (Taren Point, NSW, Australia). Trifluoroacetic acid was
purchased from Sigma-Aldrich and formic acid was purchased from Hopkin and
Williams (Essex, UK). Crude morphine and other opiate samples were supplied by
Sun Pharmaceutical Industries Australia Pty (Port Fairy, Victoria, Australia) that
were extracted from Papaver somniferum following proprietary protocols. All solid
samples were prepared at a concentration of 10 mg mL-1 in water prior to LC/MS
analysis and liquid samples were diluted by a factor of 10 with 10% acetic acid
(VWR International, Fontenay-sous-Bois, France).
Standards were made up into a 10 mL acetonitrile stock with 10 mg each of
thiourea, caffeine, phenol, anthracene and 10 µL of dimethyl phthalate and n-propyl
benzene (purchased from Sigma-Aldrich Pty Ltd, Castle Hill, NSW, Australia).
The standard stock was then added (50 µL) to 6 different compositions of
acetonitrile (50 µL, 200 µL, 400 µL, 500 µL and 600 µL) making up 1 mL with
water. Viscosity was adjusted to address solvent mismatch by adding an appropriate
amount of either isopropyl alcohol or ethanol (Chem-Supply Pty Ltd, Gillman SA,
Australia) according to viscosity calculations (Table 2.3).
Liquid Chromatography Mass Spectrometry (LC/MS)
Mass spectrometry analysis was performed with an Agilent Technologies
6210 LC/MSD TOF system (Agilent Technologies, Mulgrave, Victoria, Australia).
It was equipped with an Agilent technologies 1200 series solvent degasser, binary
pump and an Agilent technologies 1100 series autosampler (ALS). Positive ion
polarity was used in conjunction with nitrogen drying gas (5 L min-1) and nitrogen
nebuliser gas (30 psi). The morphine sample was injected (5 µL) into a linear
gradient over 40 minutes with an initial mobile phase of 5% ACN (95% water) that
increased to 100% ACN. All analyses were run at a flow rate of 0.5 mL min-1.
Chromatographic data were obtained and processed with Agilent ChemStation
software. The sixteen columns that comprised the stationary phase library for this
study are listed in Table 2.1.
29
Table 2.1. Details of the HPLC columns utilised.
BRAND TYPE SIZE DIMENSIONS PART NUMBER
Column 1 Agilent Poroshell 120 EC-C8 2.7 µm 4.6 ×100 mm 695975-906
Column 2 Agilent Poroshell 120 EC-CN 2.7 µm 4.6 × 100 mm 695975-905
Column 3 Phenomenex Kinetex PFP 100 Å 2.6 µm 4.6 × 100 mm 00D-4477-EO
Column 4 Phenomenex Kinetex Phenyl-Hexyl 100 Å 2.6 µm 4.6 × 100 mm 00D-4495-EO
Column 5 Phenomenex Synergi Fusion - RP 80 Å 4 µm 4.6 × 150 mm 00F-4424-EO
Column 6 Phenomenex Synergi Hydro - RP 80 Å 4 µm 4.6 × 150 mm 00F-4375-EO
Column 7 Phenomenex Luna C5 100 Å 5 µm 4.6 × 50 mm 00B-4043-EO
Column 8 Phenomenex Luna HILIC 200 Å 5 µm 4.6 × 150 mm 00F-4450-EO
Column 9 Cosmosil Cosmosil πNAP 2.5 µm 4.6 × 100 mm 08084-51
Column 10 Cosmosil Cosmosil 5PBB-R 5 µm 4.6 × 150 mm 05697-21
Column 11 Cosmosil Cosmosil 5NPE 5 µm 4.6 × 150 mm 37904-01
Column 12 Agilent Poroshell 120 EC-C18 2.7 µm 4.6 × 100 mm 695975-902
Column 13 Agilent Poroshell HPH-C18 2.7 µm 4.6 × 100 mm 695975-702
Column 14 Agilent Poroshell 120,SB-C18 2.7 µm 4.6 × 100 mm 685975-902
Column 15 Agilent Poroshell 120 Bonus RP 2.7 µm 4.6 × 100 mm 695968-901
Column 16 Agilent Pursuit XRs 3 Diphenyl 3 µm 4.6 × 100 mm A6021100X046
30
Two-Dimensional High Performance Liquid Chromatography
An Agilent Technologies (Mulgrave, Vic., Australia) 1290 2D
chromatograph was used for all separations, including: an auto sampler; solvent
degasser; capillary pump; binary pump; quaternary pump; column thermostat; 2
position 8 port switching valve; and two diode array detectors (one in each
dimension, both of which monitored absorbance at 210 nm and 254 nm). The 2D-
HPLC separations were performed using 100 µL injections and a controlled
temperature of 25°C. All mobile phases contained 0.1 % formic acid. The first
dimension was completed at a flow rate of 0.1 mL min-1 with an initial mobile phase
of 15% ACN (85% water) that increased to 40% ACN over 40 minutes, increasing
to 60% ACN over a further 20 minutes and was then isocratic (60% ACN) for a
further 30 minutes. A counter gradient was used in the first dimension, after the
column, with an initial mobile phase of 9% ACN (91% water) to 4% ACN over the
first 40 minutes, decreasing to 0% ACN over the next 20 minutes and was then
isocratic at 100% water for a further 30 minutes at a flow rate of 0.5 mL min-1. The
second dimension was completed at 2 mL min-1 with an initial mobile phase of 10%
ACN (90% water) that increased to 15% ACN over 2 minutes, increasing to 60%
ACN over a further minute. Thus, the comprehensive two-dimensional separation
had an overall completion time of 90 minutes. Two-dimensional HPLC was
completed on-line with a modulation time of 4 minutes, whereby a fraction volume
of 75 µL was transferred to the second dimension via a sample loop (500 µL) and
switching valve that was controlled with the HPLC control software. It should be
noted that ACN was used throughout as due to solvent mismatch it is of significance
to two dimensional chromatography
Data analysis
The chromatographic data obtained from the LC/MS was analysed with
OpenMS® 1.11.1 [162, 163]. The OpenMS® pipeline included a file converter, file
filter, resampler, noise filter (Savitzky Golay), peak picker wavelet, feature finder
(metabo) (metabolite features), collect, feature linker (unlabelled QT) and a text
exporter. Output files were present at the feature finder, feature linker and the text
exporter. All parameters within these tools were default except in the file filter
where m/z values were specified to be within the range of 50-700 and intensity
31
values were set to a range 1,000-30,000,000 counts. The output files from
OpenMS® consisted of a list of retention times for features that were common
amongst the investigated HPLC columns. This data was processed with Wolfram
Mathematica 10 (distributed by SHI UK, Milton Keynes, England) to match these
retention times and predict the relative retention behaviour using algorithms written
in-house to inform the analyst of possible orthogonal columns. Two-dimensional
contour plots were prepared with Wolfram Mathematica 10.
32
Results and discussion
Identifying the columns that provide the greatest separation space utilisation
for 2D-HPLC can be a labour intensive process, particularly when comparing
column combinations individually [150]. Previous researchers have attempted to
shorten the process of identifying orthogonal columns however issues such as
differing peak elution times on the variety of columns tested and unknown sample
matrices [66] has prompted the use of LC/MS to manage the peaks using molecular
masses which overcomes these issues and allows for an accurate comparison of the
columns. When analysing a complex sample matrix, where the components are
completely unknown, using standards to aid in column selection is not possible [66]
as they may not accurately represent the unknown compounds from a structural
perspective.
OpenMS® was used here to identify the columns that achieved a high
degree of fcoverage in a 2D-HPLC separation of an opium poppy extract. Although
this was not its intended purpose, it allowed a large variety of columns (256
combinations) to be compared in a significantly reduced time relative to the
traditional trial-and-error process. Additional benefits were obtained with a
reduction in method development costs, volume of solvent needed and amount of
sample required. Method development expenditure is a significant factor in
industry, where a more efficient outcome reduces the time and wage of employees
which is a major cost to the company.
OpenMS® was able to help process the data from each column based on
the masses obtained from the actual samples. The information pipeline that was
used by OpenMS® in this work is illustrated in Figure 2.2, which outlines the data
processing pipeline utilised to identify the most orthogonal columns.
33
Figure: 2.2: The final OpenMS® pipeline applied to map peaks of common mass between HPLC columns.
The initial part of the pipeline (highlighted in red) was used to analyse
samples for each of the different columns individually; this included a file
converter, resampler, noise filter, peak picker, and feature finder. In order for
OpenMS® to be able to read the mass spectrometry data files, they were converted
from the file format .mzdata to .mzml. The customised pipeline then employed a
file filter where parameters such as intensity, retention times and m/z values were
optimised to focus on the areas of interest [162]. These parameters were optimised
through the TOPPView feature of OpenMS® as shown in Figure 2.3. For this
pipeline, a set of rules were developed to selectively extract the mass values (50 to
700) from each of the input files. The LC/MS map was then transformed into a re-
sampled map before a noise filter was employed. The noise filter ‘Savitzky Golay’
was utilised for this data to smooth the baseline and eliminate peaks caused by noise
[163]. The peaks in each of the samples were then acquired from the mass spectra
using the ‘peak picker wavelet’, and all features found using the ‘feature finder
metabo’ [163]. In the second stage of the pipeline (shown in green) each of the
individual files were collected together. The ‘feature linker unlabelled QT’ then
grouped the corresponding features from multiple maps to produce an output file
[163]. From this pipeline, results were then statistically processed using
Mathematica and algorithms designed in house.
Input files File converter File Filter Resampler Noise Filter SGolay
Peak Picker Wavelet
Feature Finder Metabo
Feature Linker
Unlabelled QT
CollectText Exporter
Output Files Output Files Output Files
34
Figure 2.3: Thebaine sample a) before optimised parameters, b) optimised parameters (m/z values: 50-700 and intensity 1,000 – 1,000,000)
35
Table 2.2 lists the expected fcoverage for each column combination after using
OpenMS® to extract the retention times of features from LC/MS chromatographic
data and Mathematica to predict the fractional surface coverage when combined in
a hypothetical 2D-HPLC analysis. From the sixteen columns tested with CPS
morphine, the two columns that displayed the greatest predicted fcoverage were
identified as the Agilent Pursuit XRs 3 Diphenyl and the Agilent Poroshell 120 EC-
C18, with an expected fcoverage of 0.48 (Figure 2.4A). The retention mechanisms of
these stationary phases are different and should therefore operate on distinct aspects
of the analytes’ chemical structure. For example, diphenyl columns comprise of
two phenyl groups that are connected via a silica group and are suited to separate
aromatic compounds due to their strong dipole-dipole hydrogen bonding and π-π
interactions [164]. Surfaces that utilise these π-π interactions have been previously
used for the separation of opiate compounds and natural product extractions at large
[165]. To contrast the diphenyl separation mechanism, the C18 column separates
on the basis of analyte size and hydrophobicity via a chain of 18 carbons [166]. The
C18 surface is the workhorse of modern chromatography and has myriad
applications leading to its almost universal use in modern HPLC separations [167].
36
Table 2.2. Predicted fcoverage of each column combination. Refer to Table 2.1 for column identification.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 0.44 0.12 0.32 0.24 0.16 0.16 0.20 0.20 0.28 0.24 0.12 0.20 016 0.32 0.24
2 0.44 0.16 0.20 0.24 0.20 0.28 0.28 0.16 0.32 0.24 0.28 0.16 0.24 0.40 0.24
3 0.12 0.16 0.12 0.20 0.20 0.40 0.36 0.16 0.36 0.40 0.32 0.16 0.36 0.16 0.20
4 0.32 0.20 0.12 0.24 0.12 0.16 0.20 0.16 0.20 0.36 0.12 0.36 0.16 0.40 0.16
5 0.24 0.24 0.20 0.24 0.16 0.16 0.24 0.28 0.12 0.16 0.20 0.24 0.20 0.36 0.24
6 0.16 0.20 0.20 0.12 0.16 0.12 0.24 0.20 0.20 0.40 0.16 0.20 0.12 0.24 0.20
7 0.16 0.28 0.40 0.16 0.16 0.12 0.16 0.20 0.28 0.20 0.16 0.16 0.16 0.16 0.24
8 0.20 0.28 0.36 0.20 0.24 0.24 0.16 0.24 0.16 0.32 0.28 0.20 0.24 0.36 0.24
9 0.20 0.16 0.16 0.16 0.28 0.20 0.20 0.24 0.40 0.28 0.16 0.16 0.20 0.20 0.28
10 0.28 0.32 0.36 0.20 0.12 0.20 0.28 0.16 0.40 0.36 0.40 0.24 0.24 0.28 0.32
11 0.24 0.24 0.40 0.36 0.16 0.40 0.20 0.32 0.28 0.36 0.20 0.24 0.24 0.28 0.32
12 0.12 0.28 0.32 0.12 0.20 0.16 0.16 0.28 0.16 0.40 0.20 0.20 0.16 0.16 0.48
13 0.20 0.16 0.16 0.36 0.24 0.20 0.16 0.20 0.16 0.24 0.24 0.20 0.20 0.36 0.20
14 0.16 0.24 0.36 0.16 0.20 0.12 0.16 0.24 0.20 0.24 0.24 0.16 0.20 0.16 0.24
15 0.32 0.40 0.16 0.40 0.36 0.24 0.16 0.36 0.20 0.28 0.28 0.16 0.36 0.16 0.20
16 0.24 0.24 0.20 0.16 0.24 0.20 0.24 0.24 0.28 0.32 0.32 0.48 0.20 0.24 0.20
37
The two plots in Figure 2.4 demonstrate the bins used for the predicted
separation. A value of orthogonality can tell how much of the fcoverage the separation
is utilising, however it is not indicative of where the space is being used. If a
predicted plot is too correlated (diagonal trend) it indicates that the two stationary
phases are operating in a very similar manner which in turn leads to an inefficient
separation of compounds in both dimensions [158]. This could lead to a separation
only in the first dimension and thus no benefit is achieved by using a second
column. This is illustrated in Figure 2.4A and B, with the latter demonstrating a
correlated separation where the bins are all located along the diagonal and uses
minimal bins with an fcoverage of only 0.20. Figure 2.4A however has a much larger
fcoverage of 0.48 and is less correlated, thus should have a better stationary phase
separation. The difference here shows the significance behind this method as it
allows an elimination of two columns that are not orthogonal.
38
(A)
(B)
Figure 2.4. Scaled retention times of peaks predicted by OpenMS® and the associated occupied bins that have been filled in the fcoverage calculation for (A) Agilent Pursuit XRs 3
Diphenyl and Agilent Poroshell EC-C18 and (B) Cosmosil πNAP and Phenomenex Synergi Fusion.
C18 (normalised retention time)
πNAP (normalised retention time)
Syn
ergi
Fus
ion
(n
orm
alis
ed r
eten
tion
tim
e)
Dip
heny
l (n
orm
alis
ed r
eten
tion
tim
e)
39
Comprehensive online 2D-HPLC requires that the second dimension be
significantly shorter than the first, requiring optimisation of the columns selected
to obtain the fastest possible separation [149]. Due to the requirement of a quick
second dimension, the methods that were utilised to produce the LC/MS maps were
unsuitable to use for the 2D-HPLC analysis. Another consideration is solvent
mismatch from the first to the second dimension which can potentially cause
viscous fingering and a solvent front that is visible on chromatograms as band
broadening, where the peaks are shorter and wider [168] and can increase retention
times [169]. This is illustrated in Figure 2.5 where a chromatogram produced by
Mayfield et al. [170] highlights the effects of viscous fingering.
Figure 2.5: Representative chromatogram by Mayfield et al. [170] highlighting the effects of viscous fingering on a HPLC separation.
A counter gradient is another method that can address differing
compositions thus eliminating the solvent front formed by mismatched starting
compositions and as such it was investigated here during method development. A
counter gradient utilises a third pump and enters the system after the first dimension
column but before transferring to the second dimension ensuring that the fraction
matches the starting mobile phase composition in the second dimension. This
40
method developed by Stevenson et al. [171] was shown to significantly improve
the peak shape of a separation and as such was trialled here in both online and
offline with a determination that a counter gradient significantly improves the
separation by eliminating the large solvent front, which can be an issue in particular
when viscous fingering is formed [172].
Figure 2.6: Illustration of viscous fingering as determined by Catchepoole et al. in the paper “Visualising the onset of viscous fingering in chromatography columns” [168].
Viscous fingering is a mismatch in solvent viscosity that results in fingers
of the heavy solvent leaching into the less viscous solvent instead of having a linear
approach. This is illustrated in Figure 2.6, where Catchpoole et al. added dye to the
more viscous solution and injected into a clear column [168]. In an attempt to solve
this problem in an alternate way to a counter gradient, isopropanol was added to a
sample of six standards in a series of differing injection compositions. This was
based on the paper by Stevenson et al. [171] where this technique was used to
predict a counter gradient. Both ethanol and isopropanol were tested to match the
viscosity of the solvent being injected into the second dimension. These two
solvents were added in the amount calculated using kinetic viscosity (Table 2.3),
where kinetic viscosity is:
41
Equation 3
The kinetic viscosity of acetonitrile was calculated to be 0.439 mm2 s-1,
water as 0.8926 mm2 s-1, isopropanol as 2.494 mm2 s-1 and ethanol as 1.232 mm2 s-
1. These values were used to calculate the required amount of isopropanol or ethanol
to match the viscosity of the starting composition.
Table 2.3. The viscosity calculations for isopropanol and ethanol.
Acetonitrile
(%v/v)
Water (%v/v) Isopropanol
(mL)
Ethanol (mL)
0.05 0.95 0 0
0.20 0.80 0.1275 0.3423
0.30 0.70 0.2125 0.5704
0.40 0.60 0.2975 0.7986
0.50 0.50 0.3825 1.0268
0.60 0.40 0.4675 1.2549
A starting percentage of 5% acetonitrile was utilised as it is a very common
initial mobile phase composition and the values were then increased at regular
intervals until 60% acetonitrile. It was observed that the peak shape of the standards
was not significantly improved by the addition of a viscosity altering additive as
shown in Figure 2.7.
42
Figure 2.7: Separation of standards, a) sample just contains standards, b) standards containing isopropyl alcohol and c) standards containing ethanol at 0.4 % Acetonitrile,
0.6 % water and 0.2975 mL-1 isopropanol or 0.7986 mL-1 ethanol
This led to the use of a counter gradient instead as this eliminated the solvent
front issue from the first to second dimension. The counter gradient was calculated
based on the Stevenson et al. [171] paper and the need for this is illustrated in Figure
2.8B and 2.9B where a solvent front is visible in the 2D-HPLC bins and contour
plots.
To reduce back-pressure concerns when operating at an elevated flowrate
in the second dimension, the 100 mm Poroshell C18 column tested with mass
spectrometry was substituted with a 50 mm Poroshell C18. With this column
length, all components of the sample were eluted within 3 minutes. Despite the
adjustments to the second dimension, the fcoverage achieved when these two columns
were utilised was 0.43 (Figure 2.8A and 2.9A), which is only slightly less than
predicted (0.48).
0
100
200
300
400
500
600
700
800
900
1000
0 2 4 6 8 10
Absorbance (mAu)
Time (mins)
No additive
Isopropanoladditive
Ethanol additive
43
(A)
(B)
Figure 2.8. Scaled retention times of peaks identified from peak picking algorithm and occupied bins of the 2D-HPLC separation of opiate alkaloids for (A) Agilent Pursuit XRs
3 Diphenyl and Agilent Poroshell EC-C18 and (B) Cosmosil πNAP and Phenomenex Synergi Fusion.
Diphenyl(Normalised retention time)
πNAP (normalised retention time)
44
A column combination with a low fcoverage was also investigated to validate
LC/MS mapping across columns and determine the reliability of this method. The
Cosmosil πNAP and Phenomenex Synergi Fusion-RP columns were expected to
operate via similar retention mechanisms, with the πNAP having preferential
retention to aromatic compounds through π-π interactions and the Fusion-RP being
a polar embedded C18; however, after mapping the masses between these columns,
a fcoverage of 0.20 was expected (Figure 2.8B). After performing the 2D-HPLC
analysis of the opiate extraction with the πNAP column in the first dimension, the
analysis was determined to have an fcoverage of 0.34 (Figure 2.9B). The 2D-HPLC
plots for the combination of diphenyl × EC-C18 and πNAP × Synergi Fusion are
represented in Figure 2.9A and 2.9B, respectively. In the OpenMS® prediction the
components were correlated, which was also observed in the real sample; however,
the actual analysis did give a significantly higher fcoverage than predicted. This was
due to the solvent mismatch from the first dimension to the second dimension
artificially decreasing the retention time of some of the peaks [171].
45
(A)
(B)
Figure 2.9. (A) 2D-HPLC plot using an Agilent pursuit XRs 3 Diphenyl in the first dimension and a Phenomenex Poroshell 120 EC-C18 in the second dimension. (B)
Illustration of a 2D-HPLC separation with a small fcoverage: the first dimension (πNAP) was completed at a flow rate of 0.05 mL min-1 with an initial mobile phase of 15%
aqueous ACN that increased to 55% aqueous ACN over 200 minutes and was isocratic (55% aqueous ACN) for a further 10 minutes. The second dimension (Synergi fusion) was completed at 2 mL min-1 with an initial mobile phase of 20% aqueous ACN that increased
to 60% aqueous ACN over 3 minutes. Two-dimensional HPLC was completed on-line with a modulation time of 5 minutes and completion time of 210 minutes.
46
The separations were optimised with a sample from the later stages (data
above) of the extraction process in order to use a less complex opiate alkaloid
sample, so as to not overload the computing processors (4 Gig 8 core which is
typical of a standard computer as of 2017). As the previous approach showed
successful multidimensional separation of the opiates, the universal capabilities of
the methodology were tested with other more complex opiate samples (data below)
from the Sun Pharmaceutical extraction process. The opiate processing stream and
impurity standards consisted of 68 samples that have all been analysed with this
2D-HPLC method. The following two-dimensional plots (Figure 2.10 and 2.11)
confirm that the protocol developed can be applied to a variety of opiate samples
and afford a separation of the components in a complex sample. An identical
method was used when scaling up the complexity of the separation and the degree
of space utilisation was maintained.
47
a)
b)
Figure 2.10: fcoverage plots for the samples (a) G1527-030G (0.47) and (b) G1527-027C (0.47).
0.0 0.2 0.4 0.6 0.8 1.00.0
0.2
0.4
0.6
0.8
1.0
1st Dimension
2nd
Dim
ensi
on
0.0 0.2 0.4 0.6 0.8 1.00.0
0.2
0.4
0.6
0.8
1.0
1st Dimension
2nd
Dim
ensi
on
48
a)
b)
Figure 2.11. Two example 2D-HPLC chromatograms (G1527-030O and G1527-027C) representing complex samples from an opiate alkaloid extraction process line. (Please
note Purple represents minima in this example set at zero and Red maxima)
49
Table 2.4: fcoverage calculated from each stage of the processing stream
Sample Average fcoverage Sample Average fcoverage
G1527-027A 0.39 G1527-028A 0.38
G1527-027B 0.42 G1527-028B 0.33
G1527-027C 0.45 G1527-028C 0.32
G1527-027D 0.39 G1527-028D 0.34
G1527-027E 0.37 G1527-028E 0.34
G1527-027F 0.36 G1527-028F 0.40
G1527-027G 0.37 G1527-028G 0.38
G1527-027H 0.36 G1527-028H 0.44
G1527-027I 0.37 G1527-028I 0.40
G1527-027J 0.43 G1527-028J 0.38
G1527-027K 0.37 G1527-028K 0.41
G1527-027L 0.35 G1527-028L 0.43
G1527-027M 0.36 G1527-030A 0.37
G1527-027N 0.40 G1527-030B 0.40
G1527-027O 0.37 G1527-030C 0.38
G1527-027P 0.39 G1527-030D 0.39
G1527-027Q 0.39 G1527-030E 0.36
G1527-027R 0.37 G1527-030F 0.36
G1527-032A 0.41 G1527-030G 0.36
G1527-032B 0.29 G1527-030H 0.29
G1527-032C 0.32 G1527-030I 0.39
G1527-032D 0.24 G1527-030J 0.32
G1527-034A 0.29 G1527-030K 0.33
G1527-034B 0.31 G1527-030L 0.30
G1527-034C 0.41 G1527-030M 0.36
G1527-034D 0.36 G1527-030N 0.42
G1527-034E 0.31 G1527-030O 0.46
G1527-034F 0.41 G1558-131A 0.41
G1527-034G 0.39 G1558-131B 0.31
G1527-034H 0.34 G1558-131C 0.40
G1527-034I 0.32 G1558-131D 0.41
G1527-034J 0.32 G1558-131E 0.37
G1527-034K 0.39 G1558-131F 0.37
G1527-034L 0.39 G1527-034M 0.36
50
Some of these samples contain a large number of compounds whilst others
are not as complex which is noticeable from the fcoverage’s determined from each
stage of the processing stream as shown in Table 2.4. Using previously described
peak picking algorithms and the bins method, the measured average fcoverage of the
complicated extracts vary between 0.24 and 0.46 depending on the complexity of
the sample. The lowest average fcoverage achieved was 0.24, however this sample is
an impurity standard papaveraldine, where there are only a few other components
present beside the major peak. The largest average fcoverage achieved was 0.46 for
the sample G1527-030O which consists of the mother liquor after the target
compound oripavine has been precipitated out of the processing stream. The
complex sample G1527-027C is from the start of the morphine processing stream
and had the next greatest average fcoverage at 0.45. The fcoverage plots for G157-030O
and G1527-027C are illustrated in Figure 2.10a and 2.10b respectively, where it is
evident that a large amount of the available separation space is utilised and the
peaks are not focused along the diagonal line. The 2D-HPLC plots for these two
samples are also shown in Figure 2.11 where the complexity of these samples can
be observed. These plots highlight the need for a two-dimensional separation as all
the compounds would not be resolved using a single column and the methodology
developed here delivers the capacity to resolve the complex samples.
51
Conclusions
Due to developing technology, the vast quantity of available HPLC columns
has given considerable choice when selecting columns for HPLC analysis [66].
When selecting columns for LC×LC, it is essential that their retention mechanisms
are as different as possible. However, current methods to identify columns that fit
this criteria are time consuming and not ideal for blind samples. Utilisation of
LC/MS to compare columns has eliminated this problem as the unknown
compounds are compared via their molecular mass. This allows columns to be
blindly selected for unknown sample components without comparing each column
pair individually. Even when the sample is known, injecting multiple standards for
each column is still incredibly time consuming, whereas with the above method all
standards can be combined, requiring each column to only be analysed once. This
method not only allows for more columns to be compared in a shorter amount of
time, but also spares the amount of sample and mobile phase required and the man
power and wages for the time spent on the selection thus reducing the cost
excluding of course the purchase price of a mass spectrometer.
The use of the program OpenMS® enables successful selection of
orthogonal columns for LC×LC. It was determined here that a first dimension
diphenyl stationary phase and a C18 column in the second dimension was optimal
for the separation of complex opiate extraction samples. An fcoverage of 0.48 was
predicted by OpenMS® and algorithms designed in-house, with an actual
separation fcoverage of 0.43 obtained after solvent mismatch was addressed with a
counter gradient in the second dimension. The developed method was then applied
to 68 opiate processing samples where average fcoverage values ranged between 0.24
and 0.46, depending on complexity of the sample. For experiments requiring repeat
analysis, the methods outlined here could allow for a comprehensive study to find
the columns with the greatest orthogonality.
52
CHAPTER 3:
OPIATE PROCESS STREAM ANALYSIS
53
Chapter overview
The industrial process involved in extracting opiates from the opium poppy
is quite complex and involves many steps. Of interest to Sun Pharmaceuticals is
what components are present in each stage of the process, and which areas afford
the most amount of chemical change in the manufacturing line. This chapter
focuses on the analysis of the processing stream and the development of the
platform technology to chemically map the manufacturing process. Principal
Components Analysis (PCA) is employed after 2D-HPLC analysis of 65 samples
from morphine, thebaine and oripavine streams. Mass spectrometry is then used to
identify the masses that the PCA process identifies as the largest areas of change.
The platform technology developed here enables the analysis of each stage of the
manufacturing process or changes between different batches to be identified.
54
Introduction
An important aspect of manufacturing pharmaceuticals is the elimination of
impurities from the final product. Impurities can come in a variety of forms; as a
starting material (including solvents and reagents), as a product of the
manufacturing process or as a degradation product [71]. Impurities in
pharmaceutical products are an issue that is closely monitored with strict
regulations followed, as impurities could alter the effectiveness and safety of the
product [173].
Sun Pharmaceuticals extract opiates for use in medications, making it
essential that the impurity levels are low. If impurities are high the batch fails,
costing money and time, which has a negative impact on the business. Buyers of
the opiates want to know the product they are getting is pure, leading to enquiries
about unknown trace impurities. Of great interest to the manufacturers of
pharmaceuticals is where the impurities come from and if there is a way to prevent
them from forming or a way to eliminate them early in the process to ensure they
don’t end up in the final product [174]. With improvements in the detection limits
afforded by modern analytical chemistry instruments manufacturers can now
determine impurities in ever diminishing concentrations. Due to having a better
understanding of the composition of the products manufactures need to develop
strategies to mitigate the impact of the impurities in the material produced.
The use of a combination of 2D-HPLC, PCA and LC/MS has the potential
to allow mapping of the opiate processing streams and discern what is changing at
each stage. Principal components analysis (PCA) affords a deeper understanding of
what is occurring in samples. This technique has previously been utilised in the
chemical manufacturing of wine to inform vintners of the relationship between
wine quality and the grapes physiochemical index [175]. It has also been used for
the quality control of a medication in Belgium, where counterfeit pharmaceuticals
are an issue. PCA was used to fingerprint medicine based on impurities and other
components from the manufacturing process to compare legal and counterfeit
profiles [176]. This highlights that this technique is very useful in chemical
manufacturing and quality control.
55
This is also important for different batches of poppy straw as growing
opium poppies is subject to natural growing environments and each crop is likely
to be affected in different ways. This can be caused by different geographical
locations or even different cropping techniques employed on the farm. These
factors can lead to varying levels of extracted compounds or the potential for new
or differing impurities. This was noted in a study by Hopfer et al. where wine from
different vineyards was analysed and found to differ in the elemental compositions
[177]. The winery these wines were processed in also had an effect on the elemental
composition [177]. In addition to this, new poppy hybrids are trialled and
occasionally these new breeds could have a huge impact on the amount of opiates
and or new impurities. In cases such as this, the platform technology developed
here has the potential to make the analysis of any fundamental changes easier and
allow monitoring of batches and crops before processing.
56
Experimental
Chemicals and samples
Water was obtained from a Milli-Q filtration system (Merck Millipore,
Victoria, Australia). HPLC grade acetonitrile (ACN) was purchased from Ajax
Finechem Pty. Ltd. (Taren Point, NSW, Australia) and formic acid added at 0.1%
to all mobile phases, was purchased from Hopkin & Williams Ltd, Essex, UK.
Opiate processing stream samples were supplied by Sun Pharmaceutical Industries
Australia Pty (Port Fairy, Victoria, Australia) that were extracted from Papaver
somniferum following proprietary protocols. All solid samples were prepared at a
concentration of 10 mg mL-1 in water prior to 2D-HPLC analysis and liquid
samples were diluted by a factor of 10 with 10% acetic acid (VWR International,
Fontenay-sous-Bois, France). All solid samples were prepared at a concentration
of 1 mg mL-1 in water prior to LC/MS analysis and liquid samples were diluted by
a factor of 100 with 10% aqueous acetic acid. Where these liquid processing
samples fit into the process is illustrated in Figures 3.2 and 3.3. Other samples not
in these diagrams include impurity standards and powdered samples. Because of
the sensitive proprietary nature of the work, codes were used as-is to de-identify
the samples to make publication of the data possible while not divulging proprietary
information.
Two-Dimensional High Performance Liquid Chromatography
This was performed on the same instrument with the same method
developed and outlined in Chapter 2.
Principal Components Analysis (PCA)
The 2D-HPLC separations were exported from the chromatography data
system ChemStation as .AIA files before the peak information was extracted by
first smoothing the raw data. Algorithms built in-house in the Wolfram
Mathematica environment were used for all data analysis. The first and second
derivatives of the curve were then scrutinised to find the start, end and maxima
times for resolved, overlapping and shouldering peaks. After this the peaks were
mathematically described and their area found by applying a combination of model
57
Gaussian curves to the data as illustrated in Figure 3.1 where Pap et al. [178] fitted
model Gaussian peaks to a GC separation. To ensure the same compound is not
spread over several first dimension fractions the peak profiles are compared in
adjacent chromatograms.
Figure 3.1: A representation of the fitting of model Gaussian curves to data by Pap et al. [178] for a GC separation of n-penthane and n-hexane.
An algorithm designed in-house [65] using Wolfram Mathematica 10
(distributed by SHI UK, Milton Keynes, England), processed all chromatograms in
batch using the below peak picking parameters:
Least squares tuning parameter (λ): 1 × 106
Initial standard deviation for Gaussian fitting: 0.03 (min)
Savitsky-Golay window size: 3; polynomial order: 2
Minimum peak response threshold: 0.2
First derivative minimum response threshold: 5 × 10-4
Second derivative minimum response threshold: 1 × 10-5
90% overlap requirement to group peaks in adjacent
chromatograms.
58
Data pre-treatment was then performed in Mathematica 10 on the extracted
2D-HPLC fingerprints with an in-house developed algorithm. In short this included
peaks being reported as absolute multidimensional elution times after their
retention times were analysed before peak retention times were binned at 6 second
increments and the peak areas summed to define variables. Normalisation was
conducted on these variables for chosen pairs of sample sets and the mean
subtracted before a batch wise calculation of the first 2 principal components of the
sample set pairs. Reconstruction of the multidimensional retention times for the
loadings led to the development of the scores and loadings plots from principal
components analysis using a covariance matrix. This two dimensional data was
transposed to the bubble plot formation and all loadings were represented by the
colours which can be used to either identify the positive and negative loadings or
relate to the sample of interest. Throughout the text below this is described in each
of the figure legends respectively
Liquid Chromatography/ Mass Spectrometry (LC/MS)
A Shimadzu LCMS-8040 Triple Quadrupole Liquid Chromatograph Mass
Spectrometer (LC-MS/MS) was used for the analysis. The data in chapter 3 and
chapter 4 were generated in MS/MS all other data was generated in single
dimension MS. Positive ion polarity was used in conjunction with nitrogen drying
gas (15 L min-1) and nitrogen nebuliser gas (3 L min-1). All processing stream
samples (Chapter 3) were analysed with an injection volume of 1 µL and a flow
rate of 0.5 mL min-1 for both methods. Method A (scaled to match first dimension
LC×LC) was completed with an Agilent pursuit XRs 3 diphenyl column at an initial
mobile phase of 15% ACN (85% water) that increased to 40% ACN over 8 minutes,
increasing to 60% ACN over a further 4 minutes and was then isocratic (60% ACN)
for a further 6 minutes. Method B (scaled to match second dimension LC×LC) was
completed with a Phenomenex Poroshell C18 column at an initial mobile phase of
10% ACN (90% water) that increased to 15% ACN over 8 minutes, increasing to
60% ACN over a further 4 minutes and was then returned back to 10% ACN for a
further 4 minutes. Chromatographic data were obtained and processed with
59
Shimadzu profiling solutions software with total ion count thresholds set to either
1,000,000, 500,000 or 1,000 counts.
60
Results and discussion
Figure 3.2 is an illustration highlighting where in the system the 18 samples
obtained from Sun Pharmaceutical Industries morphine processing stream
originated. Samples included here comprise of those from the main opiate line, the
final product and the solvent waste streams. Likewise, Figure 3.3 illustrates the
oripavine and thebaine manufacturing streams, where selected waste stream
samples and the main processing line for both opiates have been provided for
analysis.
61
Figure 3.2: Diagram of the samples from the morphine processing stream.
G1527-027A
G1527-027D
G1527-027F
G1527-027E
G1527-027B
G1527-027H Fresh Solvent
G1527-027K
G1527-027JG1527-027I
G1527-027M
G1527-027L
G1527-027R
G1527-027Q
G1527-027P
G1527-027N
G1527-027O
G1527-027G CPS
Crystallise & precipitate
First aqueous (Aq.) extract
Aq. - tank 2 pH adjustment
Aq. – Post Filtration pH adjustment
Aq. Added complexing agent with impurity removal
Aq. extract tank 1
PSE (partway through extraction)
Aq. input
Aq. waste
Solvent waste
Fresh aq.
Rich solvent
Rich aq. composite
Rich aq. (CE) fresh
Solvent
AEX aq.
62
Figure 3.3: Diagram of the samples from the thebaine and oripavine processing stream.
G1527-028A
G1527-028L
G1527-028E
G1527-028B
G1527-028D
G1527-030H
G1527-030F
G1527-028HG1527-030A
G1527-028F
G1527-028C
G1527-030D
G1527-028G
G1527-028J
G1527-030B
G1527-030I
G1527-030G
G1527-030C
G1527-030L
G1527-030E
G1527-030J
G1527-030M
Oripavine
Precipitate
G1527-030K
G1527-030O
Process continues –No samples
First Aq./ water soluble solvent extract
Heated & water soluble solvent removed
Aq. Filtered (Drum)
Rich solvent
Steam stripped of solventSpent Aq. freshFresh solvent
Rich solvent
Bulk solvent (rich) Thebaine, No oripavine
Rich solvent Caustic
Once extracted solvent (PST)
Rich aq. Oripavine (CEO)
Waste aq.
Waste aq.
Twice extracted solvent (PST 2)
Waste solvent
Acid
Acid wash
Acid
Caustic
Caustic
Acid wash
Aq. Rich extract 2
Aq. Rich extract 1
63
Principal Components Analysis (PCA)
Raw chromatographic data requires peak alignment before chemometric
analysis can be performed due to minor variations in retention times. To align the
2D-HPLC peaks from this data correlated optimised warping [179] was trialled,
however it proved to be inefficient on a standard computer where it was unable to
align 6 chromatograms with approximately 216,000 data points. In order to achieve
a more rapid data analysis that is able to be performed by a standard computer, the
data set was reduced by extracting the multidimensional retention times [65] and
cumulative peak areas [180] to create characteristic sample fingerprints.
Keeping the restrictions used to determine peak parameters constant, batch
PCA was performed on comparisons of 89 sample pairs, each with triplicate 2D-
HPLC data. In this chapter two samples (G1527-028H and G1527-030O) have been
selected as representative of all sample sets to show the developmental process of
the techniques employed. A more in depth analysis of other samples of interest is
included in the case studies in Chapter 4 where the significance of the chemistry at
each stage of the processing stream is discussed. This chapter deals with the
development of the analytical protocol. The samples selected are of interest due to
their location in the oripavine processing stream, where G1527-028H comes
directly before the target compound is removed and G1527-030O is the spent
stream after the removal of oripavine.
In Figure 3.4, the 2D-HPLC data (sample G1527-030O) has been illustrated
in two distinctive ways to highlight the value of the information that can be gained
from these plots. The most common (Figure 3.4a) way of illustrating the data is via
a contour plot, where the purple represents the background, the red signifies a large
intensity peak and the white dots indicate the peak maxima. A disadvantage of this
plotting technique is the ability to appropriately visualise the third dimension to the
data, the peak intensity. This occurs because compounds that are less-absorbing
may be lost when scaling the data or large peaks lose their detail when trying to
visualise these smaller peaks. An alternative way of visualising this data is a bubble
plot developed specifically for this project where a characteristic fingerprint was
produced by picking the key chromatographic features. The bubble plot allows the
peak height to be visualised by the diameter of the circle, where the centre is the
64
peak maxima at the corresponding first and second dimension retention times (x-
and y-axis, respectively). The diameter of the bubble in Figure 3.4b is proportional
to the relative peak area and they have been normalised between 0 and 1. It is also
important that the size of the bubble is only indicative of the height of the peak and
does not convey the resolution of the separation based on overlapping bubbles.
65
A)
B)
Figure 3.4: Illustrative 2D-HPLC chromatograms of sample G1527-030O (replicate 1). A) Contour plot with peak intensities from 1 to 75 amu is represented and is scaled from purple (low) to red (high), B) peak maxima are located at the centre of the bubbles and
the peak intensities are represented by the bubble diameter, which have been normalised.
66
The sample or variable matrix is required to be rectangular for the principal
components calculation. This required the use of the binning technique, which is
where the separation space is broken up into square ‘bins’ [155]. The information
can then be characterised by the amount of bins the data points occupy. This was
illustrated in Chapter 2, where for example Figure 2.10 shows grey squares that
represent the bins occupied by data points and the white area are those that are not
occupied. Using this method, the retention data was divided evenly in 0.1 minute
intervals for the 90 minute run time to overcome the differing number of peaks
detected in each chromatogram. By creating the rectangular sparse matrix, the
analytes with slightly differing retention times were aligned thus eliminating the
need to align the peak retention data prior to chemometric analysis. Prior to binning
the data, it was normalised and the mean from each bin’s cumulative peak area was
subtracted. Principal components analysis was completed with randomised
triplicate 2D-HPLC analysis of each sample in batch using a covariance matrix and
scores plots of the first two principal components were created along with their
corresponding loadings plots.
Figure 3.5 displays the PC1 × PC2 scores plots for the samples G1527-028H
and G1527-030O, where 67.4% of the sample variance is accounted for in PC1 and
89.4% by the 2 components. A clear distinction can be noted between the two
samples with the G1527-028H replicates grouped in the first quadrant of PC1 and
PC2 replicates of G1527-030O in the second quadrant. One replicate of G1527-
030O is not grouped tightly with the other two, however this can be explained by
this sample being of a solvent base, where some solvent remained in the sample on
the first replicate thus affecting this result. Despite this, it can be observed that there
was a large distinction present in PC1 between these two samples with a large
majority of variance covered by the first two principle components.
67
Figure 3.5: PC1 × PC2 scores plot of G1527-028H and G1527-030O samples.
Indications of what variables are responsible for the bulk variance between
samples were observed by the loadings of the principal components. For this
analysis the loadings of the first two principal components were converted into a
multi-dimensional representation of the data, which is illustrated by the bubble
plots in Figure 3.6. The centre of the bubble corresponds to the coordinates of the
multidimensional retention time while the absolute normalised loading on each
principal components equals the radius of the bubble. A larger impact on the change
that peak had on the overall discrimination between groups on the scores plot is
signified with a larger diameter bubble. To simplify, the more significant the
change in peak area, the larger the corresponding bubble). The colour of the bubbles
is also significant as they indicate which sample is responsible for an increasing or
decreasing change in average cumulative area, where blue indicates a larger
average area of the peak in G1527-028H and red indicates the peak in G1527-030O
had a larger average area.
68
A)
B)
Figure 3.6: PCA loadings of A) PC1 and B) PC2 for the samples G1527-028H and G1527-030O. A blue bubble indicates a larger average area in the 028H samples, and a
red bubble indicates the 030O samples were more concentrated.
69
The comparison of these two samples indicated from the loadings of PC1
and PC2 that the most variation between samples occurs in the lower left hand
section of the multidimensional separation. The high concentration compounds
would most likely be highly polar in the aqueous streams, but as the pH is adjusted,
polar compounds may become neutral and crystallise out. As such they would
demonstrate the most significant change in relative area. In each of the loadings
plots, the bubbles were normalised separately so the PC1 and PC2 changes cannot
be directly compared, for example the largest bubble in the second principal
component does not necessarily mean that there was a greater impact than a smaller
first principal component and vice versa. The biggest change in PC2 was caused by
a peak eluting at 72 minutes in the first dimension and 3.17 minutes in the second
dimension. The data for this peak (72, 3.17 minutes) of the triplicate
chromatograms of G1527-030O are presented in Table 3.1, in addition to the
loadings of this peak when compared to G1527-028H. The loadings values of both
principal components were similar, but this smaller value was more significant
compared to the other loadings in PC2 compared to PC1. Due to this peak having
a greater influence on PC2 in the scores plot it was likely that the relatively large
value of the standard deviation (σ = 2.289) of these areas was causing most of the
spread on this axis in Figure 3.6.
Table 3.1. Replicate fingerprint data for multidimensional chromatograms
of G1527-030O.
Replicate tR,1 tR,2 Area Loading PC1
(Rep 1-3)
Loading PC2
(Rep 1-3)
1 72 3.160 7.812 -0.0841 -0.06394
2 72 3.164 2.781
3 72 3.165 3.151
To find if the molecules were being formed by the manufacturing
process or whether impurities present at the end of each stage were from the opium
poppy itself, these loading plots from the 2D-HPLC fingerprints were interrogated.
Representative chromatograms of before (G1527-028H) and after (G1527-030O)
the key compound oripavine is extracted are illustrated in the contour plots in
70
Figure 3.7, where the PC1 loadings are overlayed. It was postulated that if
differences between these samples were from the poppy biomass extraction alone,
the resultant changes in these chromatograms would be peaks that are missing from
G1527-030O that remain in G1527-028H, which would be represented by yellow
bubbles only. However, the biggest changes between these two chromatograms
were the presence of new compounds throughout the G1527-030O chromatogram
as highlighted by green bubbles in Figure 3.7. The G1527-030O sample was taken
from a reaction vessel that had been under storage for several days between cycles
and thus kept at atmospheric temperature and pressure. This would suggest that
new compounds seen in this sample are a result of degradation or oxidation
reactions taking place in the vessel and thus confirm that changes are taking place
during the manufacturing process. To date, Sun Pharmaceuticals have done very
little work in this space and the work presented in this thesis has enabled the
company to further refine its priorities for sample interrogation.
71
A)
B)
Figure 3.7: Overlay of PC2 loadings on 2D-HPLC chromatograms of A) G1527-028H and B)G1527- 030O (replicate 1). A yellow bubble indicates a larger average area in the G1527-028H samples, and a green bubble indicates the G1527-030O samples were more
concentrated.
72
An indication that this protocol is working well can be determined by the
larger yellow disk observed in Figure 3.7 from the G1527-028H sample, this
identifies the removal of oripavine before the next sample (G1527-030O) the key
step at this point of the process. This reiterates some features that were expected
however other aspects such as the formation of new compounds in the latter sample
are an important finding in the manufacturing process. This is significant for quality
control and to understand what is occurring at each step of the process. If new
compounds are formed at stages such as this and then become problem impurities
in the final product, it is possible that it can be addressed by not allowing the batch
to be exposed for as long. To enable the identification of the compounds causing
major changes in between processing steps, LC/MS can be utilised to pinpoint the
responsible masses.
Mass Spectrometry
Mass spectrometry was utilised to develop an in depth understanding of the
peaks that were responsible for the most amount of change between process
samples. It was anticipated that between G1527-028H and G1527-030O there
would be a significant amount of change to the amount of oripavine present due to
the step in between these two samples being the removal of this compound. To fully
validate this, a standard of oripavine was also analysed by 2D-HPLC and LC/MS
as shown in Figure 3.8. Due to the significantly greater sensitivity of the mass
spectrometer in comparison to the UV-vis detector, the oripavine was too
concentrated in all three samples. Thus, when the mass spectrometry data is
observed the change in oripavine was not noted as a big change as in all three
samples the detector recorded a maximum total ion count for this compound. This
however is significant as it highlights that although the bulk of oripavine is
precipitated out at this step, there is still a large portion carried over to the mother
liquors that continue on in the process.
73
Figure 3.8: 2D-HPLC contour plot of a supplied oripavine sample (G158-131A).
Whilst oripavine is expected to change, other unknown components of the
sample also differ, requiring the use of a combination of 2D-HPLC, PCA and mass
spectrometry to identify those compounds. The areas responsible for the greatest
change are presented in Table 3.2, in conjunction with the first and second
dimension retention times. These retention times can then be matched to the scaled
LC/MS data to locate what compounds correspond to these changes and the signal
intensities can then be compared to identify the mass of the responsible molecule.
74
Table 3.2. The PCA loadings (Figure 3.5A) scores for the comparison of
G1527-028H and G1527-030O.
First dimension RT Second Dimension
RT
PC1 loadings
scores
24 0.85 0.243382
24 1.05 0.196234
28 0.65 -0.126360
36 0.65 -0.147570
32 1.05 -0.458970
32 0.85 -0.492530
28 0.55 -1.000000
The positive PC1 loadings scores indicate the compound has a greater
average area in G1527-028H than in G1527-030O and the negative values indicate
a greater presence in the latter, as visualised in Figure 3.5. From these values and
coordinates the LC/MS data was consulted to identify the masses responsible for
the changes (Table 3.3).
75
Table 3.3: Masses corresponding to first dimension retention times
identified by PCA (PC1).
Ion m/z Ion RT
(mins)
Adj. RT
(mins)
G1527-028H
(Ion counts)
G1527-030O
(Ion counts)
314.19 4.80 24.02 1322110 0
617.30 4.94 24.68 4101544 3714015
298.17 5.02 25.10 20000000 20000000
299.19 5.02 25.10 20000000 20000000
314.19 5.68 28.38 0 1121306
329.20 5.71 28.57 1303501 1699879
318.15 5.72 28.59 0 1462715
328.18 5.73 28.65 7626469 8626498
370.18 5.79 28.94 2270866 2851843
330.20 5.81 29.06 20000000 20000000
344.22 6.45 32.23 7411381 14269454
328.18 6.54 32.70 17465896 20000000
329.20 6.55 32.73 3215862 5032537
344.23 6.60 33.02 20000000 20000000
345.24 6.62 33.12 20000000 20000000
312.19 7.22 36.09 18173297 20000000
316.19 7.22 36.12 0 1353942
314.23 7.23 36.13 3935946 7778106
313.21 7.23 36.17 3725287 6779475
342.21 7.41 37.06 3828494 4116965
In this chapter, the methodology has been shown to be effective at
identifying key features of the data at hand. To realise the full potential of this
methodology, application to the entire processing streams of morphine, thebaine
and oripavine has been performed in Chapter 4 to establish changes in a chemical
system that encompasses numerous steps and chemical environments.
76
Conclusions
A platform technology has been developed here for the mapping of opiate
extraction streams that will allow for a greater understanding of changes along the
process. A multidimensional separation addressed the complexity of the process
samples and achieved the necessary resolving capacity not afforded by
conventional HPLC. These two-dimensional HPLC fingerprints were used in
conjunction with PCA loading plots to highlight key differences between pairs
of sample sets taken from the Sun Pharmaceuticals morphine and thebaine
extraction process. It was found that relative changes between two sets of samples
could be identified by the analyst using bubble plots of loadings overlayed on a
contour representation of the data and relative peak area could be labelled as either
increasing or decreasing. The major changes could then be pinpointed using scaled
LC/MS, to find the mass of the peaks of interest.
77
CHAPTER 4:
APPLICATION OF PLATFORM TECHNOLOGY TO MONITOR COMPONENTS OF INTEREST
78
Chapter overview
This chapter applies the methods previously described in Chapters 2 and 3
to a series of case studies based on monitoring key compounds of interest within
the opiate processing streams. Segments of the processing stream where the key
differences between the samples were observed are investigated to highlight the
impact that each stage of the process has on the chemistry of the sample matrix.
This chapter highlights the utility of this type of methodology to solve tangible
process chemistry problems and builds capacity for the process owners to make
timely informed decisions.
79
Introduction
The opiate processing stream is complex, with many steps required in order
to extract and purify opiates on a large scale. These steps can be as simple as being
held in a holding tank or as complex as large column separations. In-process control
tests at each stage of drug extraction are important to the final quality of the
intermediate or API. As such, it is important to know what is occurring along the
process in case of an abnormality or to track trace impurities to ensure the quality
of the drugs [181]. Manufacturers of pharmaceuticals must adhere to the strict
guidelines set out in the ICH-Q3A of 2006 [182] and comply to the relevant
pharmacopeia monographs. The ICH-Q3A specifically targets the regulation of
impurities in new drug substances to ensure all known and unknown drug
impurities are reported correctly and meet safe standards [182, 183]. Currently a
threshold for impurities and degradation products is in place for active
pharmaceutical ingredients (API) with a maximum of 0.1% unknown compounds
allowed [71]. However, as sensitivity of detection methods improve it has a flow
on effect to the customer of these products as they begin to question these trace
impurities, thus applying pressure to the manufacturer to know the identity of each
of these unknown compounds.
Quality critical attributes (QCA) and quality critical parameters (QCP) are
both important complimentary concepts. The quality critical attributes are
characteristics that could potentially impact the safety and effectiveness of the
product [184] and as such quality critical parameters are put into place by
companies to ensure the attributes are met. To ensure effective quality critical
parameters are in place an understanding of what may affect the impurities is
required and identification of compounds (e.g. structure) would help with this
insight.
Of further interest is where these compounds originate and what is lost and
gained at each step of the process, to determine if impurities of interest are a product
of the manufacturing stream, introduced with raw materials or solvents, or if they
come from the key starting materials, i.e. opium poppy, itself. This also has the
potential to be complex as the variety and growing conditions of the crops have the
potential to influence the type and quantity of impurities, and as such highlights the
80
importance of having robust analytical monitoring that can detect and measure
these changes. Currently there are many molecules of interest to Sun
Pharmaceutical Industries Australia that to date have only been identified by their
molecular weight. While the list is extensive, for this thesis the molecules with a
molecular weight of 285 amu, 297 amu and 622 amu have been identified as
significant impurities in their processing streams by Sun Pharmaceuticals. It is of
great interest to know where within the process stream these impurities originate
for Sun Pharmaceuticals to develop informed and subsequently targeted
elimination or mitigation protocols.
Molecule of Interest 1 (285 amu)
The molecule with a mass of 285 amu is of great interest to Sun
Pharmaceuticals as it has been observed as an impurity for a long time in the
morphine processing stream and is present in the final product. Interestingly
morphine also has a molecular mass of 285 amu and it has been proposed that it is
likely to be an isomer of morphine. Isomers are typically difficult to separate and
this may explain why it is such a persistent impurity in the morphine line.
Molecule of Interest 2 (622 amu)
The molecule with a mass of 622 amu is also a key component, more
structural elucidation including NMR spectroscopy has been performed on this
molecule and the compound has been identified by Sun Pharmaceuticals as O-
Methylsomniferene (O-M-Somniferene), an asymmetric dimer of thebaine (Figure
4.1). As such this compound may possess similar chemical attributes to thebaine
causing it to be persistent within the processing stream. This compound has
historically been a problem in the thebaine processing stream and as such it is of
great interest where this compound emanates from in order for the process owners
to limit its production or improve on its removal.
81
Figure 4.1: Structure of O-M-Somniferine.
Molecule of Interest 3 (297 amu)
A further impurity of both the thebaine and oripavine processing stream
(due to the same starting point/product) is the molecule with a molecular mass of
297 amu, previously determined to be nor-thebaine (Figure 4.2). Interestingly, this
impurity has the same molecular mass as oripavine and due to its potential as an
isomer is of great interest to the company. Further, nor-thebaine is used as a
precursor to pharmaceutical intermediates such as those used for buprenorphine
[185]. Understanding its formation either in the plant or process may provide a
method for simple and cheap production over existing synthetic routes.
Figure 4.2: Structure of Nor-thebaine.
82
The use of in silico optimised 2D-HPLC has been shown to be an effective
way of resolving isomers and recent work, supplementary to this thesis, has shown
the application of this process to the resolution of ephedrine and pseudoephedrine
in methamphetamine seizure samples [186]. In short, this work was developed for
the interrogation of methamphetamine samples including model, real world seizure,
and laboratory synthesised samples. The protocol used Drylab® software to rapidly
identify the optimum separation conditions from a library of chromatography
columns. The optimum separation space was provided by the Phenomonex Kinetex
PFP column (first dimension) and an Agilent Poroshell 120 EC-C18 column
(second dimension). To facilitate a rapid 2D-HPLC analysis the particle packed
C18 column was replaced with a Phenomenex Onyx Monolithic C18 without
sacrificing separation performance. The Drylab® optimised and experimental
separations matched very closely, highlighting the robust nature of HPLC
simulations.
A fully comprehensive two-dimensional separation was completed with a
total analysis time of 115 minutes, the separation is illustrated in Figure 4.3. The
retention times predicted by DryLab® are shown by the white circles, which have
been overlaid on the actual two-dimensional separation (Figure 4.3).
Figure 4.3: Overlaid DryLab® and actual two-dimensional separation of the model
methamphetamine seizure sample.
83
After training the software with data from separations occurring over 2
different gradient durations the simulation was able to extrapolate 2 variables (i.e.
S and kw) for each peak that are required to predict the retention time with varying
gradient durations. DryLab® is able to simulate the retention profile of a complex
separation matrix with a single step (linear) gradient, or by adding multiple steps;
a 4 step gradient is presented here. When this process is done to optimise each
retention dimension individually a 2D-HPLC retention map can be extrapolated
and the final multidimensional simulation is predicted.
This is one of the early points that in-silico software has been used to predict
a two-dimensional separation of methamphetamine samples and indeed
forensically relevant samples at large. This is of great significance for the
development of 2D-HPLC protocols, a process that historically takes considerable
time. The total method development time for the separation presented in Figure 4.3
was less than two hours with in-silico aided optimisation. This is significantly less
than the time required for typical two-dimensional separation optimisation that
normally requires many individual sample injections [153]. The results of 2D-
HPLC separation prediction was impressive as all components of interest were
closely matched (Figure 4.3). This is not a trivial outcome due to the complex
nature of the separation mechanisms involved allowing the user to have high
confidence in the in-silico prediction. The power of 2D-HPLC observed in this
work is directly applicable to the challenge of identifying key components in
complex process chemistry samples. The work in this chapter takes this technology
a step further by incorporating the first application of PCA to two-dimensional
chromatographic data.
84
Experimental
Case studies
The techniques previously described in Chapters 2 and 3, have been applied
to samples from select sections of the process stream. Initially, 2D-HPLC was
performed on each of the samples and compared as mentioned in Chapter 2 and
subsequently analysed using Principle Components Analysis (PCA) as described
in Chapter 3. The samples were then analysed with mass spectrometry in a method
scaled to match the first and second dimension of the 2D-HPLC, as previously
described in Chapter 3. The areas of change were identified from the PCA plots
and all mass spectral peaks within the corresponding retention times were
compared to determine which compound was the cause of the greatest changes in
between the respective steps of the process stream. Specific points of interest were
examined using these methods and reported as individual cases defined by the list
below.
Case 1: G1527-027N, G1527-027O and G1527-027G.
Case 2: G1527-027L and G1527-027M
Case 3: G1528-028C, G1527-028D and G1527-028H
85
Results and discussion
Molecules of interest
The opiate processing stream contains many molecules of interest to both
the company and its customers. Three masses of interest have been focused on in
this chapter, where they have been mapped throughout all 81 opiate processing
stream samples in the main line, waste lines and crude impurity standards. The
presence of the molecules at 285 amu, 297 amu and 622 amu are illustrated through
the processing stream in Figures 4.4 and 4.5.
86
Figure 4.4: Morphine processing streams with masses 622 (present when red) and 285 (present when blue).
G1527-027A
G1527-027D
G1527-027F
G1527-027E
G1527-027B
G1527-027H Fresh Solvent
G1527-027K
G1527-027JG1527-027I
G1527-027M
G1527-027L
G1527-027R
G1527-027Q
G1527-027P
G1527-027N
G1527-027O
G1527-027G CPS
Crystallise & precipitate
First aqueous (Aq.) extract
Aq. - tank 2 pH adjustment
Aq. – Post Filtration pH adjustment
Aq. Added complexing agent with impurity removal
Aq. extract tank 1
PSE (partway through extraction)
Aq. input
Aq. waste
Solvent waste
Fresh aq.
Rich solvent
Rich aq. composite
Rich aq. (CE) fresh
Solvent
AEX aq.
87
.
Figure 4.5: Thebaine and oripavine processing streams with masses 622 (preset when red) and 297 (present when yellow).
G1527-028A
G1527-028L
G1527-028E
G1527-028B
G1527-028D
G1527-030H
G1527-030F
G1527-028HG1527-030A
G1527-028F
G1527-028C
G1527-030D
G1527-028G
G1527-028J
G1527-030B
G1527-030I
G1527-030G
G1527-030C
G1527-030L
G1527-030E
G1527-030J
G1527-030M
Oripavine
Precipitate
G1527-030K
G1527-030O
Process continues –No samples
First Aq./ water soluble solvent extract
Heated & water soluble solvent removed
Aq. Filtered (Drum)
Rich solvent
Steam stripped of solventSpent Aq. freshFresh solvent
Rich solvent
Bulk solvent (rich) Thebaine, No oripavine
Rich solvent Caustic
Once extracted solvent (PST)
Rich aq. Oripavine (CEO)
Waste aq.
Waste aq.
Twice extracted solvent (PST 2)
Waste solvent
Acid
Acid wash
Acid
Caustic
Caustic
Acid wash
Aq. Rich extract 2
Aq. Rich extract 1
88
Molecule of interest 1 (285 amu)
Morphine has a molecular mass of 285, however there are isobaric
components present throughout the processing stream that have the same mass and
are considered impurities. One such compound of this mass is of great interest as it
is present in the final morphine product and to date the structure is unknown. The
company are interested in whether the peak was an impurity extracted from the
opium poppy or if it is a product of the processing stream, to determine if it is
something that could be controlled or eliminated. From the LC/MS data of all the
process samples, the impurity was found in the first extraction (G1527-027A) and
continued through the main stream to the CPS morphine (Figure 4.4). This indicates
that the steps taken along the process have not been effective in removing this
impurity. The powdered samples, CPS-M (G1558-131D) and technical morphine
(G1558-131B) were both tested where the latter is a purified form at 94% compared
to 79%. The impurity with a molecular weight of 285 amu was only present in the
CPS morphine (G1558-131D). In the thebaine processing line it was found to only
be present in two samples tested, G1527-030N and G1527-030O which are both at
points along the stream where concentration occurs leading to the possibility that
the impurity is minute in the rest of the stream but becomes noticeable via mass
spectrometry when the stream is concentrated up.
The fragmentation pattern of compounds with the same mass can be used
to elucidate the chemical structure. The two compounds at m/z 286, as shown in
Figure 4.6 were identified by mass spectrometry and have different fragmentation
patterns and chromatographic retention times that vary by approximately two
minutes. The first peak is consistent with the fragmentation pattern of morphine,
however the second is significantly different indicating that it is a structural isomer.
89
a)
b)
Figure 4.6: Mass spectral fragmentation pattern of a) morphine and b) the unknown impurity of the same mass.
50.0 75.0 100.0 125.0 150.0 175.0 200.0 225.0 250.0 275.0 m/z0.0
2.5
5.0
7.5
Inten. (x100,000)165.1
181.1152.1
128.1
115.055.0 171.0201.091.0 209.1 286.177.1
50.0 75.0 100.0 125.0 150.0 175.0 200.0 225.0 250.0 275.0 m/z0.0
0.5
1.0
1.5
2.0
2.5
Inten. (x100,000)185.1157.0
58.1
129.1193.1
229.2153.1211.4 267.2122.8
90
To further characterise this compound and the following two, NMR
spectroscopy needs to be performed however due to the complex nature of these
samples the target compounds required separation from the other components in
the sample to enable structural elucidation. This was attempted by utilising a one-
dimensional HPLC separation with the previously discussed diphenyl column on
an analytical scale. Using an Agilent fraction collector after the detector, target
peaks were collected on a time based scale, the mobile phase evaporated and then
reconstituted with deuterated methanol before NMR analysis. A portion of the
fractions was taken for direct injection mass spectrometry to confirm the collected
peaks were the target compound. Whilst this was confirmed the lack of sensitivity
of the NMR meant the peaks could not be characterised, even after in excess of 100
fractions were collected and combined. Semi-preparative HPLC was also trialled
to increase the amount of sample that could be utilised however due to poor
chromatography, this idea was also abandoned. Due to the difficulties with
collecting enough of the target compound an alternate way to concentrate the
impurity is required however this is now beyond the scope achievable here.
Molecule of interest 2 (622 amu)
The second mass of interest is 622 amu which corresponds to the compound
O-M-Somniferine (Figure 4.1). This mass was investigated as it is a known
impurity of the thebaine processing stream, however the results found with LC/MS
also identified a mass corresponding to the m/z 623 molecular ion in the morphine
samples. This is a mass not currently looked for in the morphine processing line,
however it is postulated to be an asymmetric dimer of thebaine. The MS/MS data
showed thebaine in the fragmentation pattern and as such this supported the
proposed compound assignment. As illustrated in Figure 4.4 this compound is in
the first sample extract from the poppy straw pellets and continues through the main
processing stream until the caustic wash where it is expelled into the recycled
solvent waste. This suggests that it originates in the raw product and is efficiently
extracted under the conditions afforded by the process. Finding this impurity in the
recycled solvent waste may also have implications on the process as the impurity
is being reintroduced into the stream towards the start of the process, potentially
leading to an increased concentration throughout the batches. This highlights the
importance of tracking impurities as this is potentially a sampling point that can be
91
monitored to ensure the concentration of the impurity is not increasing significantly
and possibly causing issues in the extraction process.
This compound is also observed in the thebaine stream, only emerging at
G1527-030A, the rich solvent after thebaine and oripavine are separated (Figure
4.5). This compound stays in the stream despite caustic impurity washes and is
removed at the point of the acid washes. During the acid washes the compound is
absent from the main stream but appears in each of the sequential downstream acid
washes. It is postulated here that this occurs due to the main stream being
significantly more dilute than the acid wash sample, where the compound is
gradually washed into these lines.
At the first acid wash stage (G1527-O30G) that uses organic acid an isomer
of m/z 623 is observed denoted by two peaks of the same mass with different
retention times. This highlights the importance of the concentration steps as the
second isomer is only observable when concentrated downstream. As such for any
thorough investigations into these compounds that Sun Pharmaceuticals may need
to do for a client, concentration of the sample streams prior to the acid wash would
be required. Alternatively, this isomer could be forming at this stage due to a
chemical reaction induced by a lowering of the pH.
At the third acid wash stage using mineral acid (G1527-O30K) the original
isomer of m/z 623 was no longer observed. This indicates that the efficiency of the
organic acid wash was sufficient to remove the primary m/z 623 isomer. Further,
the mineral acid washes completely remove the second m/z 623 isomer from the
process stream. This work highlights the issues associated with compounds of the
same mass that are observed only at certain points in the stream and this technology
allows process owners to carefully monitor these compounds.
Molecule of interest 3 (297 amu)
The third mass of significant interest is 297 amu, which corresponds to the
impurity nor-thebaine (Figure 4.2). This compound is a known impurity in the
thebaine stream and shares the same mass as oripavine and a further unknown
compound present primarily in the morphine stream. Due to its presence in the
combined thebaine and oripavine stream, this impurity is of particular interest. The
impurity was not detected in the morphine stream, however its presence was traced
92
through the thebaine and oripavine stream as illustrated in Figure 4.5 denoted by
yellow highlights.
This impurity elutes at 6.9 minutes and as such elutes after oripavine. From
this the impurity is first present at sample point G1527-028L, after the opiates have
been extracted from the straw with water miscible solvent and subsequently heated
to remove the solvent. The impurity is then present in the main processing stream
until the acid washes. During the organic acid treatment the impurity is present in
both the washes and the process stream. At the mineral acid stage the remainder of
the compound is extracted out. In the sample G1527-030G, this compound is found
to be the most concentrated, exceeding the detection capabilities of the LC/MS.
This sample is the first acid wash to form aqueous rich extract and as such is more
concentrated. This compound is also found in the rich aqueous oripavine and the
resulting mother liquor. This shows the capacity to gain information on quality
from the concentration stages of the processing streams.
Case 1: G1527-027N, G1527-027O and G1527-027G.
Case 1 is from the morphine processing stream and focuses on the samples
G1527-027N, G1527-027O and G1527-027G. G1527-027N (see Figure 4.4) is
fresh rich aqueous caustic extract. G1527-027O directly follows G1527-027N and
is the rich aqueous composite. This is then subjected to a crystallisation and
precipitation step, where CPS morphine is recovered and G1527-027G is the
mother liquor left from this process (Figure 4.4).
The PC1 × PC2 scores plots for G1527-027N and G1527-027O were first
consulted to determine the sample variance. Figure 4.7 illustrates that 63.5 % of the
sample variance is accounted for in PC1 and 92.3 % by the first two components.
93
Figure 4.7: PC1 × PC2 scores plot of G1527-027N and G1527-027O samples.
The loadings plot for G1527-027N and G1527-027O were compared and
the changes investigated. It was found that there were significant changes between
these two samples with eight major differences. This is illustrated in Figure 4.8 and
the retention times of these changes are shown in Table 4.1.
Figure 4.8: PC1 loadings plot for G1527-027N against G1527-027O from 2D HPLC analysis.
94
Table 4.1: The top eight PC1 loading scores for G1527-027N and G1527-027O
First Dimension RT
Second Dimension RT
PC1 Loadings score
24 0.5 1.000 28 0.8 0.479 28 1.0 0.466 32 3.3 0.172 24 0.6 0.172 72 3.1 0.165 32 0.6 0.144 80 3.3 0.105
These retention times were then compared to the profiling solutions LC/MS
data, where an ion count threshold of 1,000,000 was utilised to focus on the
important large changes. It is important to note that the focus here was placed on
the larger changes in order to test the protocol and that for broad application
individualised assessment of the mass spectral thresholds will need to be made. As
the technology developed here gets greater use an informed peak count threshold
will be required however the threshold described here was shown to be sufficient.
Care will need to be taken for application of this to other sample types beyond
opiates which may possess lower mass spectral ionisation efficiencies. The changes
are highlighted in Table 4.2
Table 4.2: Pseudoolecular ion mass to charge ratios, total ion counts and retention
times for the peaks with significant changes observed in the loadings plots for
G1527-027N and G1527-027O respectively (1,000,000 threshold).
Ion m/z Ion RT (mins)
Adj. RT (mins)
G1527-027O (Ion counts)
G1527-027N (Ion counts)
344.23 5.69 28.44 4129200 2329157
329.20 5.71 28.57 1171678 0
328.18 5.73 28.65 5010879 5270546
194.10 6.41 32.04 1507660 0
328.18 6.54 32.70 19225078 14066945
329.20 6.55 32.73 4228279 2858543
314.22 6.58 32.88 8123023 5099602
338.36 16.06 80.31 1408922 1634005
95
The changes at a retention time of 28.44 and 28.57 minutes (Figure 4.8)
correspond to the molecular ions m/z 344 and m/z 329 respectively. The mass to
charge ratio is consistent with the compound laudanine (Figure 4.9) which is readily
known in the opiate steam as it is a biosynthetic intermediate [187] and as such this
was compared to its impurity standard (G1527-034M), an in house fractionated
sample of laudanine (Figure 4.9). These standards are routinely separated and
concentrated within the Sun Pharmaceuticals R and D laboratory at Port Fairy. The
bespoke nature of these materials makes it difficult to find commercial standards
and as such it is important to note that these standards typically vary in their
impurity levels and care needs to be taken when using them. Interestingly the
laudanine standard was found not to match the retention time of 28.44 minutes
appearing approximately 5 minutes earlier. This implies that the molecule is likely
to be similar to laudanine including its interaction with the diphenyl stationary
phase.
Figure 4.9: Structure of Laudanine.
As mentioned previously, an ion count threshold of 1,000,000 was set, in
this study to limit the data size to a reasonable and workable amount, however if
necessary the threshold may need to be adjusted to enable a thorough investigation
of the changes of interest to other sample sets. To visualise the implications of a
significant adjustment to the ion count threshold a study was performed on the same
data set as presented in Figure 4.8 with a new threshold of 500,000 ion counts. The
molecular masses of the compounds responsible for the major changes are
presented in Table 4.3.
96
Table 4.3: Molecular ion mass to charge ratios, total ion counts and
retention times with the peaks with the most observable change in PC1 loadings for
G1527-027N and G1527-027O (500,000 threshold)
Ion m/z Ion RT (mins)
Adj. RT (mins)
G1527-027O (Ion counts)
G1527-027N (Ion counts)
344.22 5.68 28.40 4129200 2329157 345.24 5.69 28.46 998658 0 329.20 5.72 28.60 1171678 1139468 328.18 5.73 28.65 5010879 5270546 194.10 6.41 32.04 1507660 1044668 328.18 6.54 32.70 19225078 14066945 329.20 6.54 32.72 4228279 2858543 360.20 6.56 32.82 862555 772804 314.22 6.57 32.86 8123023 5099602 330.21 6.57 32.86 746301 0 315.21 6.59 32.93 1962637 996321 171.14 14.40 72.00 543210 0 228.21 14.43 72.17 693820 0 540.44 14.50 72.47 790429 680697 334.21 14.50 72.48 0 582340 304.22 14.54 72.71 856586 1004728 378.26 14.57 72.86 753073 785251 228.20 14.58 72.92 568811 0 436.29 16.09 80.44 694525 625891
As can be observed from Table 4.3, the change in threshold allowed changes
at 72 minutes to be identified. This highlights the importance of adjusting the
threshold to allow a better understanding of changes if no masses are noted at the
retention times identified by PCA analysis.
The next sampling point of interest to case study 1 is G1527-027G, which
directly follows G1527-027O (Figure 4.4). Figure 4.10 describes the loadings plot
for the samples G1527-027G and G1527-027O. As can be noted from the plot a
large amount of change can be observed between 16 and 17 minutes in the first
dimension. This is also observed in the PCA loadings scores in Table 4.4, where
the seven significant changes in the samples are shown with the biggest change
observed at a retention time of 16 minutes.
97
Figure 4.10: PC1 loadings plot for G1527-027G against G1527-027O from 2D HPLC analysis.
Table 4.4: The seven PC1 loading scores for G1527-027G and G1527-027O
First Dimension RT
Second dimension RT
PC1 loadings score
16 0.4 1.000000 16 0.3 0.779080 36 2.7 0.166485 16 0.5 0.102417 32 0.5 0.081450 56 0.3 0.076003 24 1.1 0.070652
The mass spectral peaks were compared to the known retention time
differences noting that this was performed at an ion count threshold of 1,000,000.
The mass spectral peaks that were identified at these retention times are shown in
Table 4.5, where the values highlighted in yellow are those that produced the most
significant change at that respective retention time.
98
Table 4.5: Molecular ion mass to charge ratios, total ion counts and
retention times with the peaks with the most observable change in PC1 loadings for
G1527-027G and G1527-027O. The highlighted rows indicate the values
associated with the most notable changes corresponding to table 4.4.
Ion m/z Ion RT Adjusted RT
Ion Count G1527-027G
Ion Count G1527-027O
324.11 3.29 16.45 0 6651041 286.16 3.31 16.55 20000000 20000000 308.18 3.31 16.56 0 20000000 325.13 3.33 16.63 0 1421907 669.25 3.33 16.63 3807454 0 268.14 3.33 16.64 2387128 19796688 287.17 3.33 16.64 20000000 19350048 593.27 3.33 16.64 0 2963348 288.20 3.33 16.66 8296774 12787382 192.12 3.34 16.70 20000000 0 193.16 3.34 16.72 8179070 0 316.18 3.36 16.78 1174037 0 302.18 3.36 16.79 19904082 0 603.30 3.36 16.79 11272815 0 604.20 3.36 16.79 4500447 0 240.16 3.36 16.80 7938151 4955320 318.16 3.39 16.94 7439495 0 208.13 3.39 16.96 3024467 0 314.19 4.80 24.02 1429778 0 490.25 4.87 24.34 1725044 0 194.10 6.41 32.04 6184907 1507660 344.22 6.45 32.23 5324913 0 328.18 6.54 32.70 20000000 19225078 329.20 6.55 32.73 4971785 4228279 360.18 6.55 32.77 7421490 0 378.18 6.56 32.81 2050925 0 314.22 6.58 32.88 20000000 8123023 316.19 7.22 36.12 1276425 0 314.23 7.23 36.13 2151253 0 329.23 7.23 36.17 9909379 2319867 328.21 7.24 36.18 20000000 12676532 274.29 11.31 56.53 19983254 0 275.31 11.32 56.60 3660635 0 346.31 11.33 56.65 1199210 0 230.28 11.38 56.88 7978807 0
From Table 4.5, it is clear that the three changes at 16 minutes in the first
dimension are due to the molecular ions m/z 308, m/z 192 and m/z 302. The m/z
302.18 is characteristic of 10-Hydroxymorphine, where it is present in the G1527-
99
027G sample point but is absent at the G1527-027O sample point. As the G1527-
027G sample point directly follows G1527-027O the lack of this mass in the data
is possibly due to oxidation or other reactions of compounds within the sample over
time. 10-Hydroxymorphine is a degradation product of morphine [188] leading to
the possibility that it further degrades under the right conditions to form a different
impurity. As the morphine and other unknown compounds are removed through
crystallisation, new compounds now constitute a significantly higher percentage of
the remaining liquors and due to this the large change in the presence of 10-
Hydroxymorphine can be more easily observed. This is consistent with unpublished
data from Sun Pharmaceuticals where under high pH an increase in three as yet
unknown impurities has been observed [174]. This again highlights how this type
of comprehensive sample analysis can identify key changes and can enable the
company to prioritise points of interest for detailed investigation.
This is supported by the second dimension separation where m/z 302.18 is
at an adjusted retention time of 0.288 minutes, which is consistent with the PCA
loadings plot identifying the second biggest change at 16 minutes in the first
dimension and 0.3 minutes in the second dimension. Also consistent is the intensity
change, whilst in the first dimension this mass was absent in G1527-027O, the
second dimension had a relatively small intensity (4697069 counts) indicating that
the mass is present at a much lower concentration than in G1527-027G (19633025
counts) again a direct observation of the successful elimination of a range of
impurities at this point of the morphine processing stream.
Importantly several other key components have been identified via this
process which may be targets for further study by Sun Pharmaceuticals when
defining future issues for customers. The change at 24 minutes is from the
molecular ions m/z 490, the change at 32 minutes m/z 314, the change at 36 minutes
m/z 328 and the change at 56 minutes from m/z 274. After discussion with Sun
Pharmaceutical scientists the molecular ion m/z 314 has been postulated to be
codeine methyl ether which is commonly observed in these extracts. Codeine
methyl ether has previously been reported for use to make thebaine with high yields
(83 %) in comparison to other potential starting products making it an important
alkaloid with prospective commercial viability [25].
100
Case 2: G1527-027L and G1527-027M
Case 2 is also from the morphine processing stream however the focus here
is on the solvent waste procedure. Sample points G1527-027L and G1527-027M
are both partially spent solvent and it is important to note they come from two
different Karr reciprocating-plate extraction columns that are used in this stream
before the next step and are expected to contain similar components. An example
of a Karr reciprocating column is shown in Figure 4.11 as described by Smith et al.
[189]. It is a solvent extraction column that utilises plates that move in a
reciprocating fashion to separate and purify components [190] and as such has a
wide variety of applications in chemical industries, including pharmaceutical and
petrochemical [189]. The difference between the two columns is that the first
column (sample point G1527-027L) has fresh caustic running through it whereas
the second column (sample point G1527-027M) utilises the partially rich caustic
from the first column. The sample G1527-027Q is next in line downstream from
these and is then followed by G1527-027R, which is recycled in the process at
sample G1527-027D (Figure 4.4).
Figure 4.11: Depiction of a Karr reciprocating plate-extraction column as described by Smith et al. [189].
101
As can be noted from the loadings plot in Figure 4.12, there is only one
major change between G1527-027L and G1527-027M and approximately five
minor changes denoted by the circle diameter. The PC1 loadings values were
investigated to find the coordinate retention times as displayed in Table 4.6. From
this plot the major chemical changes can be observed and correlated to the mass
spectral data.
Figure 4.12: PC1 loadings plot for comparison of G1527-027L and G1527-027M.
Table 4.6: PC1 loadings values for G1527-027L and G1527-027M.
First dimension
RT Second dimension RT PC1 loadings score
44 2.7 1.000000
44 1.5 0.119075
44 1.4 0.102645
32 2.9 0.064433
44 1.8 0.020958
44 1.6 0.016224
102
The main change between the two large scale processing columns was
interrogated with LC/MS to determine the mass of the compound which was
determined to correspond to m/z 171.14. When the 2D-HPLC plots were
consulted, as shown in Figure 4.13 the change between the two samples by eye is
small highlighting the importance of this type of PCA data handling. This
protocol has allowed the biggest changes between the samples to be found and
scales them from 0 to 1, which is an important aspect to be aware of as in the case
of these two samples the change is small.
103
Figure 4.13: 2D-HPLC contour plots of the samples G1527-027L and G1527-027M.
104
The ion m/z 171.14 has not previously been identified as a key component
of interest to Sun Pharmaceuticals. Here for the first time the company has the
capacity to assess compounds based on the PCA data processing which may afford
a deeper understanding of the components that play key roles in differentiation
between samples. This highlights the significance of the work developed here as it
could be applied to monitoring key low concentration components in batch to batch
processing samples to help inform process owners.
Case 3: G1528-028C, G1527-028D and G1527-028H
This case is from the thebaine and oripavine processing stream and is at the
point where they are separated from each other. Sample point G1527-028C is the
rich solvent containing both thebaine and oripavine, they are then separated with a
caustic wash before G1527-028D the thebaine only stream and G1527-028H the
purely oripavine stream. The 2D-HPLC contour plots for these samples are shown
below in Figure 4.14.
105
Figure 4.14: 2D-HPLC contour plots for G1527-028C and G1527-028D.
106
As highlighted by the contour plots (Figure 4.14) there are minimal small
changes observed between these two samples, however importantly there is one
large change which is due to the removal of oripavine from the stream. This was
also confirmed by the mass spectral data that showed the key change was due to
m/z 298 the molecular ion of oripavine which has been extracted. The second
largest observable change was m/z 344 which was confirmed with the use of the
standard to be the molecular ion of laudanine. This again highlights the complexity
of real world process chemistry samples and that isomers can be challenging to
monitor. Here we have shown that we can differentiate between key compounds of
the same mass and as such add to the capacity for Sun Pharmaceuticals to
interrogate samples more clearly.
This clear data set could be easily used as a visual inspection tool by the
process owners to quickly assess the quality of the separation of thebaine and
oripavine. In this case PCA data analysis would only be required to understand the
significance of the minor components within this sample. The components of low
concentration observed here are likely to be compounds that prefer solvent
conditions in the oripavine stream over the thebaine stream. This will be primarily
driven by the pH adjustment afforded by the caustic wash step and could be a key
point in the process where Sun Pharmaceuticals could focus on for efficiency
improvements.
The differences between sample points G1527-028C and G1527-028H
were then investigated to see what followed in the oripavine stream, where the
biggest change is expected to be the removal of thebaine. The contour plots were
first compared to identify the obvious changes in the two samples. Figure 4.15
shows the contour plot for sample point G1527-028H (rich in oripavine), and when
compared to G1527-028C (Figure 4.14A) the removal of the large peak associated
with thebaine at 40 minutes (First Dimension) and 2.4 minutes (Second Dimension)
can be observed.
107
Figure 4.15: Contour plot for G1527-028H.
The PC1 and PC2 loadings plots and scores were then consulted to
determine other changes that may be significant to the process. It was important for
the PCA that the samples were of the same concentration to eliminate a big change
being identified due to differing concentrations. Due to this G1527-028H was
diluted by a further factor of 10 to bring it into line with G1527-028C and G1527-
028D before PCA was applied. As can be noted in the PC1 loadings plot of G1527-
028C and G1527-028H (Figure 4.16), the biggest change identified is the removal
of oripavine.
108
Figure 4.16: PC1 loadings plot for the comparison of G1527-028C and G1527-028H.
Figure 4.17: PC2 loadings plot for the comparison of G1527-028C and G1527-028H.Showing the comparison between the two samples
109
The location of the major change (removal of thebaine) in Figure 4.16 is at
40 minutes in first dimension and 2.65 minutes in the second dimension. The next
biggest change is observed at 48 minutes in the first dimension and 2.75 minutes in
the second dimension and was identified as the previously mentioned molecular
ion m/z 171, highlighting the need for further investigation into this molecule that
is beyond the scope of this project.
This case highlights the importance of this protocol in processing samples
of increased complexity. It was clear from the 2D-HPLC contour plots that this data
was more complex than those previously investigated and it has been shown that
the key compounds of interest can be identified. Importantly minor constituents
have been observed to contribute to the loading plots and these compounds can now
be tabulated into data sets that could be applied for future use by Sun
Pharmaceuticals. The key use of these data sets will be to have a focus for any
future queries on these types of samples and gives Sun Pharmaceuticals some
capacity to pre-empt areas of concern/interest to clients.
110
Conclusions
Three masses of interest have been explored and traced through the
processing stream to determine if they originate from the opium poppy itself or as
a product of the extraction and isolation process. All three masses (unknown
impurity MW285, Nor-thebaine and O-M-Somniferine) were identified in the first
sample extracts and as such were deemed to have originated in the opium poppy
itself before being extracted alongside the target compounds.
The methods described in Chapter 2 and 3 have successfully been applied
to determine the differences between points of interest on the manufacturing
stream. Three points of interest from the opiate processing stream were investigated
and using a combination of 2D-HPLC, PCA and LC/MS differences could be
elucidated. The concept was proven to be effective by focusing on a major known
change, in which PCA identified this as a large area of difference on the two-
dimensional plots and was confirmed with mass spectrometry taking into account
the chromatographic variations observed between the two instruments. These
techniques provide Sun Pharmaceuticals with the means to identify large areas of
change between stages of the process or batches and to pinpoint the associated
masses. This is a tool that can be utilised as a quality control method for the
company or to pinpoint potential impurities.
111
CHAPTER 5:
IDENTIFICATION AND SELECTIVE DETECTION OF SYNTHETIC CANNABINOIDS
112
Chapter overview
This chapter investigates the isolation of synthetic cannabinoids from herbal
substrates followed by a systematic study of selective detection methods for these
compounds. Several herbal blends purchased in Victoria, Australia before the
recent ban are explored in order to detect and determine the synthetic cannabinoid
compounds present. Numerous standards were also synthesised at Deakin
University to fully test the detection methods on a variety of different compounds.
Chemiluminescence was used to detect these compounds with three different
reagents, of which two proved successful for the detection of all but two of the
thirteen synthetic cannabinoids. Limits of detection were calculated with both
reagents for all standards and an estimated concentration of the cannabinoids on the
corresponding herbal substrates was determined. A solid phase extraction
selectivity study was also conducted, with the use of 6 different sorbents in order
to improve the isolation of the cannabinoid from the chemically complex herbal
substrates.
113
Introduction
Synthetic cannabinoids have been abused since the early 2000s, and have
become of increasing concern to law enforcement agencies and health
professionals. These drugs have a vast array of side effects and as highlighted in
Chapter one (Figure 1.3), are responsible for many disastrous effects for users of
these herbal alternatives [40, 42]. There are a large number of known synthetic
cannabinoids and new analogues are constantly entering the market [191]. New
synthetic cannabinoids have been produced by varying the (usually heterocyclic)
core, bioisosteric replacement and/or changing or adding functional groups, often
with the intention of creating derivatives that were not prohibited by the law [44,
192, 193].
Currently to determine these compounds the samples are first extracted
using organic solvents (e.g., methanol, ethanol or acetonitrile) to exploit the non
polar nature of the synthetic cannabinoid, typically at a ratio of 1 mL of the solvent
to 10 mg of dried plant material [37, 194]. Generally, the sample extracts are then
either sonicated or vortexed and centrifuged before filtering with a 0.45 µm filter
[38, 112, 195]. The sample extracts are then analysed and the compounds detected
and identified with a variety of different techniques such as GC and LC mass
spectrometry [54].
In previous work, mass spectrometry was used to identify the actives
present in seven of the twelve synthetic cannabinoid samples utilised in this study
[196]. This technique for the identification of synthetic cannabinoids has been well
documented in the literature where it is either coupled to liquid chromatography or
more commonly gas chromatography [54].
Chemiluminescence is a sensitive and selective mode of detection, easily
coupled to HPLC, which has been exploited for the determination of numerous
compounds of forensic importance [55, 79]. Rasanen and co-workers have used
gas-phase chemiluminescence nitrogen detection for designer drugs including
some synthetic cannabinoids [197], but to date, liquid-phase chemiluminescence
has not been reported for the detection of these compounds.
Electrochemiluminescence (ECL) is a technique that has also not previously been
114
reported for use with synthetic cannabinoids. Herein, the complementary selectivity
derived from solvent extraction, chromatographic separation and
chemiluminescence detection for the determination of synthetic cannabinoids in
complex herbal sample matrices is explored.
Solid Phase Extraction
Due to the complexity of the synthetic cannabinoid samples, removing
unnecessary components from the samples will aid ‘at scene’ detection. A common
method of sample clean-up, solid phase extraction (SPE) is a selective and rapid
preparation method used for isolation, concentration, clean up and medium
exchange [198]. The aim of SPE is to separate the analytes of interest from the other
components in the sample before chromatography or analytical methods [198], and
has been used on the synthetic cannabinoid UR-144 for extraction from the herbal
product. Kavanagh et al. [61] utilised a C-18 SPE cartridge for this which is a
cartridge that has also been used for synthetic cannabinoids in urine [199, 200].
Whilst SPE is not often used to clean up herbal extractions it is commonly
used for biological samples where it is an important purification and concentration
step before mass spectrometry analysis [54]. SPE has also been used for large
variety of cannabinoids in differing biological samples such as urine [201, 202],
blood [203] and oral fluid [35, 204, 205]. Urine studies have included extracting
and quantifying the metabolites of the synthetic cannabinoids JWH‐018 and JWH‐
073 with a strong cation exchange SPE cartridge [206].
115
Experimental
Chemicals, samples and standards
Twelve commercially available herbal synthetic cannabinoid samples
(brands: Puff, Red Dot, Bombay Blue - pineapple flavour, Cloud 9, Code Black -
strawberry flavour, Northern Lights - Special K, Malibu, Atomic Bomb, Stoner,
Storm, Supanova and Voodoo; Figure 5.1) were obtained in 2013/2014 pre-ban
from a range of stores in Victoria, Australia. Synthetic cannabinoids were extracted
by adding 10 mg of the herbal mixture to 1 mL of the extraction solvent and
sonicated for 10 minutes. The organic solvents used in this study were acetone,
methanol, ethyl acetate (Chem-Supply, SA, Australia) and acetonitrile (Ajax Fine
Chemicals, NSW, Australia); all solvents were of analytical grade or higher.
Figure 5.1: Synthetic cannabinoid brands utilised in this study.
Seven synthetic cannabinoid standards (AM-2201, AB-001, 5F-AKB-48,
5F-AB-001, JWH-018, JWH-073 and 5F-AB-PINACA) were synthesized at
Deakin University. A further two synthetic cannabinoid standards (PB-22 and 5F-
PB-22) were obtained from the National Measurement Institute (NSW, Australia).
Stock solutions of these were initially prepared at 1 mM using HPLC grade
methanol (Merck, Vic, Australia) and diluted as required.
116
High Performance Liquid Chromatography (HPLC)
An Agilent Technologies 1260 HPLC with an auto injector, solvent degasser,
quaternary pump, column thermostat (20°C) and a DAD ( = 254 nm) and
chemiluminescence detector were used. Chemiluminescence reagents were
pumped to the custom-made detector with a Gilson Minipuls peristaltic pump at 1
mL min-1 as described by Adcock et al. (Figure 5.2) [207]. Reagent and column
eluent were merged at a T-piece shortly prior to entering the coiled tubing flow-
cell, as described by Holland et al. [208] (Figure 5.2). An Agilent Eclipse XDB‐
C18 column (4.6 mm × 15 mm, 5 μm particle diameter) was used for all separations.
The sample (20 µL) was injected onto the column with a flow rate of 1 mL min-1.
The mobile phase comprised HPLC-grade methanol (Merck) and Milli-Q water
(Merck Millipore, Victoria, Australia). In method A, the methanol was increased
from 0% to 100% over 40 minutes, and then held for 20 minutes, with 5 minute re-
equilibration time. In method B, both the methanol and deionised water contained
0.1% (v/v) formic acid; the methanol component was increased from 10% to 80%
over 10 minutes, and then to 100% over 10 minutes, with 5 minute re-equilibration
time.
Figure 5.2: Custom HPLC chemiluminescence detector setup as described by Adcock et al. [207]. Key: q = quaternary pump; v = injection valve; c = column; u = UV-visible absorbance detector or other non-destructive detector; p=peristaltic pump; d = flow-
through chemiluminescence detector.
117
Chemiluminescence detection
A 1 mM solution of acidic potassium permanganate (Chem-Supply) was
prepared in deionised water containing 1% (m/v) sodium polyphosphate (Sigma-
Aldrich), adjusted to pH 2.5 with concentrated sulfuric acid (Merck).
The stable tris(2,2′ bipyridine)ruthenium(II) perchlorate salt
([Ru(bipy)3](ClO4)2) was prepared by an aqueous metathesis reaction as described
by McDermott et al. [209]. Tris(2,2′ bipyridine)ruthenium(II) chloride
hexahydrate was purchased from Strem Chemicals (CA, USA) and the sodium
perchlorate was from Sigma-Aldrich. The [Ru(bipy)3](ClO4)2 salt (0.384 g L-1) was
dissolved in HPLC grade acetonitrile (Ajax) with 0.05 M perchloric acid (Ajax).
Lead dioxide (Fluka Analytical) was added (2 g L-1) and the reagent was filtered
in-line.
Manganese(IV) was prepared according to the method of Brown et al. [86].
Manganese dioxide was formed by reducing potassium permanganate (Chem-
Supply) with an excess of sodium formate (Sigma-Aldrich) and collecting the
precipitate under vacuum on glass filter paper (GF/A, Whatman, England). The
solid was washed three times with deionised water. The wet manganese dioxide
(approx. 1.2 g L-1) was placed in 3 M orthophosphoric acid (85% w/w, Chem-
Supply) and sonicated for 30 minutes before heating (80°C) whilst stirring for
approximately 1 hr. The solution was then cooled on ice and an iodometric titration
was performed using potassium iodide (Ajax), where the iodine released upon
reaction with manganese(IV) was titrated against sodium thiosulfate (Sigma-
Aldrich) solution, with a starch (Ajax) indicator determining the end-point. As an
enhancer 2 M Formaldehyde (37% w/w, Chem-Supply) was used, merging with
the HPLC eluent after the DAD but before entering the chemiluminescence
detector.
Electrospray mass spectrometry
Pre‐separated cannabinoid fractions were directly injected into an Agilent
Technologies 6210 LC/MSD ESI-TOF equipped with a 1200 series solvent
degasser, binary pump (methanol, 1 mL min-1) and 1100 series ALS. Positive ion
polarity was used under the following conditions: drying gas (N2), flow rate and
temperature (5 L min-1, 350°C), nebuliser gas (N2) pressure (30 psi), capillary
118
voltage 3.0 kV, vaporiser temperature 350°C, and cone voltage 60 V. MS data
acquisition and analysis was carried out using Agilent MassHunter Workstation
Acquisition for TOF/Q-TOF (version B.02.00 (B1128)) and MassHunter
Qualitative Analysis (version B.03.01).
Electrochemiluminescence (ECL) and Cyclic Voltammetry (CV)
Cyclic voltammetry (CV) was carried out with a glassy carbon working
electrode, silver wire as the reference electrode and platinum wire as the counter
electrode. Samples were prepared in dry acetonitrile with 0.1 M
Tetrabutylammonium hexafluorophosphate ([TBA][PF6], Fluka). A scan rate of 0.1
V s-1 was utilised for all samples, and each was referenced to 1 mM ferrocene.
Electrochemiluminescence spectra were collected with a photomultiplier
tube (PMT). As with the CV experiments a glassy carbon working electrode and
platinum wire counter electrode was utilised. Synthetic cannabinoid standards at 1
mM were first tested with Tetrabutylammonium hexafluorophosphate as a blank
before tris(2,2′‐bipyridine)ruthenium(II) hexafluorophosphate (Aldrich) was added
and the mixture re-tested. Herbal samples were tested as above, where the synthetic
cannabinoids (20 mg) were extracted with dry acetonitrile (2 mL).
Solid Phase Extraction (SPE)
Six solid phase extraction (SPE) cartridges, detailed in Table 5.1 were
purchased from Agilent Technologies. The SPE vacuum manifold was purchased
from Phenomenex (Lane Cove, NSW, Australia).
Table 5.1: Details of the Solid Phase Extraction cartridges utilised.
Cartridge Sorbent Sorbent Size
(mg/mL) Part Number
Bond Elut-C18 EWP Silica wide pore C18 50 mg, 1 mL 12102136
Bond Elut-C8 Silica C8 50 mg, 1 mL 12102059
Bond Elut-SAX Silica strong anion exchange 100 mg, 1 mL 12102017T
Bond Elut-SCX Silica strong cation exchange 50 mg, 1 mL 12102075
Bond Elut-Plexa PCX Polymer strong cation exchange 30 mg, 1 mL 12108301
Bond Elut-SI Natural silica 100 mg, 1 mL 12102010
119
All SPE cartridges were conditioned as described in Table 5.2, with the
exception of the natural silica cartridge that was only conditioned with the
extraction solvent. Samples were extracted at 10 mg of herb to 1 mL of solvent
(vortexed, filtered and analysed on HPLC) before being added to the conditioned
cartridge. The cartridges were then washed with 5% methanol (100% extraction
solvent for silica cartridge) before elution with 100% methanol (100% water for
silica cartridge). The first pass (sample solution), second pass (wash solution) and
the third pass (elution) were collected for HPLC analysis.
Table 5.2: Cartridge conditioning and extraction solvents.
Conditioning (1 mL each) Extraction solvent
(1 mL/10 mg)
Methanol , water Methanol
Methanol , water Acetonitrile
Acetonitrile, water Acetonitrile
Methanol , water Acetone
Acetone, water Acetone
Methanol, water Ethyl acetate
Ethyl acetate, water Ethyl acetate
120
Results and discussion
Synthetic cannabinoid extraction
Eight different solvents (methanol, ethanol, glacial acetic acid, DMSO,
DMF, acetone, chloroform, ethyl acetate) were trialled for the extraction of
cannabinoids from the herbal samples, during earlier work [196]. In that study
methanol, acetone and ethyl acetate were shown to be the preferred solvents. In this
work acetonitrile was also tested and compared to those three identified best
extraction solvents. Chromatograms for extracts from the 'Puff" sample, for
example are shown in Figure 5.3.
For each solvent extract from this sample, a large peak was observed in the
chromatogram at 37.7 minutes (identified as synthetic cannabinoid UR-144, as
discussed in the following section). The minor components were dependent on the
solvent used for extraction. Methanol is commonly used for the targeted extraction
of small polar molecules from within plant matrices [210] and it is likely that the
minor peaks observed before 25 minutes include polyphenolic compounds from the
herbal substrate. Due to the absorbance of acetone at 254 nm a large solvent front
was observed in the chromatogram when it was used for extraction. Acetonitrile
enabled extraction of large amounts of the cannabinoid without many of the
extraneous peaks seen with most other solvents and is recommended as the ideal
solvent for a targeted extraction and reduced sample complexity. Due to the
inhomogeneous nature of the herbal substrates it is difficult to directly compare the
concentration of synthetic cannabinoid extracted by the different solvents, however
all solvents have extracted a significant amount of the target compounds and as
such a clean extraction is the focus.
121
Figure 5.3: Chromatograms (HPLC with UV absorbance detection at 254 nm; separation method A) of the herbal product ‘Puff’ following extraction with: (a)
methanol, (b) acetone, (c) ethyl acetate, and (d) acetonitrile. The synthetic cannabinoid (UR-144) peak can be observed at 37.7 minutes.
0
20
40
60
80
0 10 20 30 40 50 60
Absorbance (mAu)
(a)
0
20
40
60
80
0 10 20 30 40 50 60
Absorbance (mAu)
(b)
0
20
40
60
80
0 10 20 30 40 50 60
Absorbance (m
Au)
Time (min)
(d)
122
Identification
Previously, five synthetic cannabinoids were identified in seven brands of
herbal ‘incense’ with the use of GC/MS and LC/MS as listed in Table 5.3. These
seven brands were analysed further in this study in addition to five new herbal
synthetic cannabinoid samples. Fractions of interest from chromatographic
separations were collected manually and analysed by ESI-MS to identify the
synthetic cannabinoids. A tentative assignment was made using ESI-MS, by
comparing the molecular ion peak to a database of known compounds and utilising
monographs made available online through www.SWGdrug.org [211]. The
analysis identified a further three synthetic cannabinoid compounds (Table 5.4).
The brands ‘Atomic Bomb’, ‘Stoner’ and ‘Storm’ contained the same two
compounds with m/z 359 and m/z 377, which were provisionally assigned as
cannabinoids PB-22 and 5F-PB-22. ‘Supanova’ also contained the m/z 359
compound. The assignments were confirmed by matching the HPLC retention
times with PB-22 and 5F-PB-22 standards. The brand 'Voodoo' also only contained
one synthetic cannabinoid (m/z 378) however is only tentatively assigned 5F-AEB
as there is little information currently available for this compound.
123
Table 5.3: Synthetic cannabinoids previously identified in the extracts of herbal samples.
Structure Common name(s) IUPAC name Brands
AM-2201 1-[(5-fluoropentyl)-1H-indol-3-
yl]-(naphthalen-1-yl)methanone
Red Dot
UR-144 (1-pentylindol-3-yl)-(2,2,3,3-
tetramethylcyclopropyl)methanone
Cloud 9; Code Black;
Puff; Special K
XLR-11 (1-(5-fluoropentyl)-1H-indol-3-
yl)(2,2,3,3-
tetramethylcyclopropyl)methanone
Bombay Blue; Cloud 9;
Special K
A796,260 [1-(2-morpholin-4-ylethyl]-1H-
indol-3-yl]-(2,2,3,3-
tetramethylcyclopropyl)methanone
Code Black
5F-AKB-48;
5F-APINACA
N-(adamantan-1-yl)-1-(5-
fluoropentyl)-1H-indazole-3-
carboxamide
Malibu
124
Table 5.4: Identified synthetic cannabinoids in the extracts of new herbal samples.
Structure Common name(s) IUPAC name Brands
PB-22; QUPIC 1-pentyl-1H-indole-3-carboxylic
acid 8-quinolinyl ester
Atomic Bomb; Stoner;
Storm; Supanova
5F-PB-22;
5F-QUPIC
1-pentyfluoro-1H-indole-3-
carboxylic acid 8-quinolinyl
ester
Atomic Bomb; Stoner;
Storm
5F-AEB;
5F-EMB-PINACA
ethyl (1-(5-fluoropentyl)-1H-
indazole-3-carbonyl)-L-valinate
*Voodoo
* Tentatively assigned
125
Chemiluminescence detection
Chemiluminescence detection has previously been utilised for the detection
of synthetic cannabinoids in the brands Red Dot, Puff, Cloud 9, Bombay Blue,
Special K, Code Black and Malibu [196]. From this previous study it was found
that acidic potassium permanganate didn’t emit light with the synthetic
cannabinoids unlike the natural cannabinoids [88]. However, they were found to
react with stable [Ru(bipy)3]3+ for four out of the five synthetic cannabinoids.
Under identical conditions (extraction with methanol; HPLC separation method
B), chromatograms were obtained for all nine synthetic cannabinoid standards
(Table 5.5) and the twelve herbal samples, with UV absorbance detection at 254
nm, and then chemiluminescence detection using three different reagents:
permanganate, manganese(IV) and [Ru(bipy)3]3+. The below Table (5.5) includes
the standards utilised in this study, in addition to the synthetic cannabinoid brands
in Tables 5.3 and 5.4. Typical chromatograms produced using the four modes of
detection (using the ‘Atomic Bomb’ sample as an example) are shown in Figure
5.4.
126
Table 5.5: Synthetic cannabinoid standards.
Structure Common name(s) IUPAC name
5F-AB-PINACA N-(1-amino-3-methyl-1-
oxobutan-2-yl)-1-(5-
fluoropentyl)-1H-indazole-3-
carboxamide
5F-AKB-48;
5F-APINACA
N-(adamantan-1-yl)-1-(5-
fluoropentyl)-1H-indazole-3-
carboxamide
5F-AB001 Adamantan-1-yl[1-(5-
fluoropentyl)-1H-indol-3-
yl]methanone
AB001 1-pentyl-3-(adamant-1-
oyl)indole
AM-2201 1-[(5-fluoropentyl)-1H-indol-3-
yl]-(naphthalen-1-yl)methanone
JWH-018 Naphthalen-1-yl(1-pentyl-1H-
indol-3-yl)methanone
JWH-073 1-Butyl-3-(1-naphthoyl)indole
PB-22;
QUPIC
1-pentyl-1H-indole-3-
carboxylic acid 8-quinolinyl
ester
5F-PB-22;
5F-QUPIC
1-pentyfluoro-1H-indole-3-
carboxylic acid 8-quinolinyl
ester
127
Figure 5.4: Chromatograms for the same herbal blend extract ‘Atomic Bomb’ obtained using separation method B with: (a) UV absorbance detection, and (b-d)
chemiluminescence detection using (b) acidic potassium permanganate, (c) manganese(IV), or (d) [Ru(bipy)3]3+. The two major peaks were assigned as synthetic
cannabinoids 5F-PB-22 and PB-22 (Table 5.3).
0
30
60
90
120
0 5 10 15 20
Absorbance (mAu)
(a)
0
2.5
5
7.5
10
0 5 10 15 20
Intensity (a.u.)
(b)
0
15
30
45
60
0 5 10 15 20
Intensity (a.u.)
(c)
0
20
40
60
80
0 5 10 15 20
Intensity
(a.u.)
Time (min)
(d)
128
The acidic potassium permanganate reagent was examined for the
chemiluminescence detection of synthetic cannabinoids in the herbal samples
because it was previously used to detect natural cannabinoids in industrial grade
hemp [88, 212]. Permanganate reacts with a wide variety of organic compounds
[213], but a much smaller subset elicit the characteristic red emission (max = 689
nm) at analytically useful intensities from excited manganese(II) [94, 96, 214]. This
reagent is particularly effective for the chemiluminescence detection of readily
oxidisable phenols, anilines and related compounds, but many species outside these
classes have also been found (sometimes unexpectedly) to generate an intense
emission [94, 96], including several non-phenolic indoles [215, 216]. When
screening the cannabinoid standards and herbal samples with this reagent (e.g.
Figure 5.2b), however, no chemiluminescence signal was obtained but they were
clearly detected by UV absorbance (Figure 5.4a). In contrast to the alkyl phenol
backbones of the natural cannabinoids [88], the synthetic mimics investigated here
are indole and indazole derivatives (Table 5.3, 5.4 and 5.5), and a dissimilar
chemiluminescence response was not surprising. The permanganate reagent did
detect some other components of extract, which were likely to include polyphenolic
compounds from the herbal materials and may pose analytical utility for substrate
characterisation.
The chemiluminescence of manganese(IV) also emanates from an
electronically excited manganese(II) species (λmax = 730 nm) [86], but this reagent
generates light with a much wider range of compounds than permanganate [97].
Despite both reagents forming an excited manganese(II) species, the
manganese(IV) has a higher maximum intensity (730 nm) compared to acidic
potassium permanganate (689 nm) due to the addition of the sodium polyphosphate
enhancer [79]. Non-phenolic indoles, such as tryptamine, tryptophan and
indomethacin, have been reported to elicit light with manganese(IV) [86, 217].
Unlike acidic potassium permanganate, the manganese(IV) reagent produced a
chemiluminescence signal with each of the synthetic cannabinoids identified in the
herbal samples (e.g. Figure 5.4c), with the exception of 5F-AKB-48 (both as a
standard and in the ‘Malibu’ sample). The manganese(IV) reagent also produced a
chemiluminescence signal with each of the synthesized cannabinoid standards with
the exception of 5F-AB-PINACA and the aforementioned 5F-AKB-48. The
129
reasons for the stark difference in selectivity between the two manganese-based
chemiluminescence reagents are yet to be fully elucidated, but appear to arise from
a combination of the lower oxidation state of manganese(IV) than permanganate,
the heterogeneous nature of reactions with the colloidal manganese(IV) reagent,
the concentration of acid and the enhancers added to each system [86, 218-221].
The absence of chemiluminescence from the combination of 5F-AKB-48 or 5F-
AB-PINACA and manganese(IV) can be ascribed to the significantly higher
oxidation potential of these indazole-3-carboxamides compared to the other
synthetic cannabinoids under discussion [222], which are indole-3-carbonyl
(ketone or ester) compounds.
The [Ru(bipy)3]3+ reagent is generally an effective chemiluminescence
reagent for the detection of aliphatic tertiary amines [92] and although a variety of
other nitrogen containing compounds also produce light with this reagent, the
emission intensity is often substantially lower for those with fewer N-alkyl
substituent's, incorporating mesomeric or resonance stabilisation, or with electron
withdrawing groups in close vicinity to the nitrogen [223]. Nevertheless, some
aromatic nitrogen-containing heterocyclic compounds that do not contain
peripheral amines, including indoles such as tryptophan [224] and indomethacin
[225], have been reported to produce chemiluminescence upon reaction with
[Ru(bipy)3]3+ at intensities sufficient for analytical applications. An appraisal of the
herbal samples with the [Ru(bipy)3]3+ reagent (e.g. Figure 5.4d) showed detectable
responses for all of the synthetic cannabinoids except 5F-AKB-48 (present in the
‘Malibu’ sample) and 5F-AB-PINACA, similar to the results obtained with the
manganese(IV) reagent.
The standard and herbal extraction of 5F-AKB-48 and the standard 5F-AB-
PINACA both failed to give a chemiluminescence reaction with any of the reagents,
however a quenching was observed with the manganese(IV) reagent (Figure 5.5)
The other synthetic cannabinoid intensities varied with the different reagents where
neither [Ru(bipy)3]3+ nor manganese(IV) proved to be superior to the other (Table
5.6). Normalised chromatograms (Figure 5.6) highlight that the difference in
intensity is greatly dependant on the analyte, as is which reagent is more effective.
For example the standard PB-22 elicits a greater signal with [Ru(bipy)3]3+ (Figure
5.6a) but the standard AB-001 has a greater response with manganese(IV) (Figure
130
5.6b). [Ru(bipy)3]3+ however, doesn't require an enhancer like manganese (IV) and
as such would be a more applicable and user friendly reagent for an at-scene
detection test.
Figure 5.5: Chromatogram of the standard 5F-AKB-48 with UV detection (green) and Manganese(IV) (brown).
Table 5.6. Comparison of chromatographic peak areas for cannabinoid
standards with chemiluminescence detection (green tick > 115, orange tick 0-115,
red cross < 0). The 115 scale has been established to allow for clear differentiation
of the components.
Cannabinoid
Peak area (a. u.)
[Ru(bipy)3]3+ Mn(IV)
PB-22 5F-PB-22 AM-2201
5F-AKB-48 AB-001
5F-AB-001 5F-AB-PINACA
JWH-018 JWH-073
59
60
61
62
63
64
65
66
67
68
0
20
40
60
80
100
120
140
160
0 5 10 15 20
Intensity (mAu)
Absorbance (mAu)
Time (mins)
131
Figure 5.6: Normalised chromatograms of the synthetic cannabinoid standards PB-22 (Blue) and AB-001 (Red) with [Ru(bipy)3]3+ and Manganese(IV) detection respectively.
55
60
65
70
75
80
85
90
95
100
0 5 10 15 20
Itensity (mAu)
Time (mins)
55
60
65
70
75
80
85
90
95
100
0 5 10 15 20
Intensity (mAu)
Time (mins)
132
Mobile phase was also found to have an impact on the signal produced by
[Ru(bipy)3]3+. These particular analytes emit a larger signal with a methanol mobile
phase in comparison to an acetonitrile mobile phase (Figure 5.7) The manganese
reagents also require a methanol mobile phase as acetonitrile quenches the signal
[94].
Figure 5.7: [Ru(bipy)3]3+ comparison of acetonitrile and methanol mobile phase for the standard 5F-AB-001.
Solutions of the synthetic cannabinoids standards were used to generate
calibration curves using HPLC with UV absorbance (254 nm) and/or
chemiluminescence detection, and the resulting figures of merit are shown in Table
5.7. It is important to note some calibration equations are linear while others are
quadratic. The intensities for the quadratic equations fell into the non-linear
chemiluminescence calibration curve range and as such a second order polynomial
was fitted to accurately find the limits of detection. Non-linear calibration models
are often required when an extended calibration range is utilised [226, 227].
50
55
60
65
70
75
80
0 5 10 15 20
Intensity (mAU)
Time (mins)
Acetonitrile
Methanol
133
Table 5.7: Analytical figures of merit
Analyte Detection RT (mins) Calibration Equation R2 LOD (M)
5F-AB-PINACA 254nm 11.90 y = 1.27 × 106x - 11.2 0.9990 9 × 10-6
5F-AKB48 254nm 16.81 y = 3. × 106x – 84.2 0.9944 9 × 10-4
5F-AB001 254nm 16.69 y = 9.86 × 105x - 0.8 0.9987 8 × 10-7
Mn(IV) 16.67 y = 1.98 × 105x + 29.9 0.9946 8 × 10-5
Ru(bipy)23+ 16.82 y = 41,114.78x + 0.1 0.9952 1 × 10-6
AB001 254nm 18.85 y = 1.44 × 106x – 0.4 0.9995 3 × 10-7
Mn(IV) 18.91 y = 5,342.59x + 1.6 0.9857 3 × 10-5
Ru(bipy)23+ 19.04 y = 7.23 × 105x + 1.7 0.9921 9 × 10-6
AM2201 254nm 14.18 y = 9.92 × 106x – 40.5 0.9999 3 × 10-6
Mn(IV) 14.36 y = -2.03 × 108x2 + 5.35 × 105x + 42.9 0.9992 1 × 10-5
Ru(bipy)23+ 14.36 y = 1.72 × 105x + 18.5 0.9937 5 × 10-5
JWH-018 254nm 16.23 y = 1.53 × 106x -1.2 0.9991 8 × 10-7
Mn(IV) 16.43 y = 130,085.71x + 9.4 0.9979 3 × 10-5
Ru(bipy)23+ 16.37 y = -3.86 × 109x2 + 1.47 × 105x + 0.1 1.0000 4 × 10-7
JWH-073 254nm 15.25 y = 8.77 × 105x - 1.0 0.9971 1 × 10-6
Mn(IV) 15.23 y = 51,200.00x + 22.8 0.9953 5 × 10-5
Ru(bipy)23+ 15.39 y = -2.44 × 109x2 + 1.10 × 105x + 0.1 1.0000 9 × 10-7
PB-22 254nm 14.83 y = 8.25 × 106x + 0.6 0.9998 4 × 10-7
Mn(IV) 14.97 y = -9.53 × 109x2 + 1.77 × 106x + 3.4 0.9993 6 × 10-6
Ru(bipy)23+ 15.06 y = 8.15 × 106x + 2.2 0.9980 1 × 10-6
5F-PB-22 254nm 13.21 y = 1.08 × 107x + 23.8 0.9953 1 × 10-7
Mn(IV) 13.35 y = -8.57 × 109x2 + 1.87 × 106x + 5.7 0.9979 3 × 10-6
Ru(bipy)23+ 13.43 y = 7.51 × 106x + 1.9 0.9975 1 × 10-6
134
From these calibrations, the total cannabinoid content extracted with 1 mL
methanol from 10 mg of herbal product was calculated. Methanol was used as an
extraction solvent here for consistency with standards and due its use as a mobile
phase. The synthetic cannabinoid content was found to range from 0.036 mg to 0.77
mg per 10 mg of product (Table 5.8). It can be noted that there is considerable
variation in the amount of cannabinoid material on the substrate. Whilst these are
only representative of a small selection of all brands that were and are available,
this highlights the dangers associated with the use of these products.
Table 5.8: Concentration of synthetic cannabinoid on herbal substrates extractions
Brand Synthetic
cannabinoid
Concentration
(M)
Concentration
(mg/mL)
Red Dot AM-2201 6.9 × 10-4 0.250
Malibu 5F-AKB-48 2.0 × 10-3 0.770
Atomic bomb PB-22 1.3 × 10-4 0.047
5F-PB-22 9.5 × 10-5 0.036
Storm PB-22 9.9 × 10-4 0.360
5F-PB-22 3.3 × 10-4 0.130
Stoner PB-22 1.9 × 10-4 0.068
5F-PB-22 1.5 × 10-4 0.057
Supanova PB-22 2.3 × 10-4 0.082
UV adsorption at 254 nm generally offers a more sensitive detection as
illustrated in Table 5.7. However chemiluminescence detection affords superior
selectivity for these samples. The limits of detection determined for the two
chemiluminescence reagents are similar and hence either could be used for the
analytes, however [Ru(bipy)3]3+ is a more user friendly reagent as it doesn't require
a formaldehyde enhancer and extra equipment for analysis [86].
Electrochemiluminescence (ECL) and Cyclic Voltammetry (CV)
Electrochemiluminescence involves the use of an applied potential to
induce a light emitting reaction, which also has the possibility to be used in an at-
scene detection device. Due to the lack of light obtained from the synthetic
135
cannabinoid 5F-AKB-48, the electrochemical properties of the analytes in solution
were investigated using cyclic voltammetry (CV). The oxidation potential can be
measured using CV, which in turn can help to understand why a compound elicits
light with chemiluminescence. Cyclic voltammetry was carried out on seven
compounds, to determine their oxidation potentials.
Figure 5.8: Cyclic voltammagram (representative of the electron transfer in the system) for the synthetic cannabinoid standard 5F-AKB-48 with reference to ferrocene.
The synthetic cannabinoids tested included PB-22 and 5F-PB-22, both of
which produced a significant amount of light when reacted with [Ru(bipy)3]3+ ,AM-
2201 of which produced a smaller amount of light and 5F-AKB-48 which produced
no light. Also tested was a precursor (Figure 5.9A) and an analogue (Figure 5.9B)
of 5F-AKB-48, both of which also did not produce light with chemiluminescence
(Table 5.9).
‐1.00E‐05
‐5.00E‐06
0.00E+00
5.00E‐06
1.00E‐05
1.50E‐05
2.00E‐05
2.50E‐05
3.00E‐05
3.50E‐05
4.00E‐05
‐0.6 ‐0.1 0.4 0.9 1.4
Current (A)
Potential (V)
136
A) B)
Figure 5.9: Structures of additional compounds tested to understand the chemiluminescence results; (a) 5F-AKB 48 Precursor and (b) 5F-AKB 48 analogue.
Tris(2,2′‐bipyridine)ruthenium(II) itself was also tested to find the
oxidation potential as was codeine, which is known to produce a significant amount
of light with this co-reactant. Table (5.10) depicts the oxidation potentials with
reference to ferrocene.
Table 5.9: Comparison of chromatographic peaks areas for additional
cannabinoid standards with chemiluminescence detection.
Cannabinoid Peak area (a.u.)
[Ru(bipy)3]3+ Mn(IV)
5F-AKB-48 Precursor 0 0
5F-AKB-48 Analogue -24 0
Table 5.10: CV oxidation potentials with reference to ferrocene.
Sample Concentration (M) Oxidation cf FC
Ru(bpy)32+ 1.0 × 10-3 0.89
5F-AKB-48 analogue 1.0 × 10-3 1.18
5F-AKB-48 Precursor 1.0 × 10-3 1.50
5F-AKB-48 2.5 × 10-4 1.40
AM-2201 1.0 × 10-3 1.16, 1. 52
PB-22 1.0 × 10-3 1.08, 1.25
5F-PB-22 2.5 × 10-4 1.08, 1.24
Codeine 1.0 × 10-3 0.64, 1.03
137
Interestingly, the oxidation potentials of the synthetic cannabinoids are
higher than the oxidation potential of Ru(bipy)32+. It has previously been thought
that the Ru(bipy)33+ reagent only produces light with compounds with a lower
oxidation potential to itself (0.89), such as codeine. However the cannabinoids with
oxidation potentials 1.16 and below produced chemiluminescence and those above
1.18 produced no emission with tris(2,2′‐bipyridine)ruthenium(II). This is
important as it sets an oxidation potential limit to these compounds for
chemiluminescence detection with tris(2,2′‐bipyridine)ruthenium(II). Some of the
synthetic cannabinoids have two oxidation potentials indicating they are oxidised
at one potential and then further oxidised at a higher potential. It is of interest that
the compounds that have two oxidation potentials are those that produce light with
chemiluminescence.
Electrochemiluminescence (ECL) was then trialled to determine which
compounds produced light with Ru(bipy)33+ when a potential was applied.
Interestingly, cannabinoid 5F-AKB-48, as well as its precursor and analogue
(which did not exhibit emission with chemiluminescence) were able to produce
light with ECL, as can be seen in Figure 5.10. In contrast, the standards which could
be measured with chemiluminescence did not afford signals with ECL. This
indicates that with the use of both methods combined, all synthetic cannabinoid
compounds tested in this study could be selectively detected in a portable manner.
This technique was also carried out for the synthetic cannabinoid herbal
samples, where extractions were performed with dry acetonitrile to determine if
this would be a viable at-scene detection method. Six of the synthetic cannabinoid
brands (Bombay Blue, Cloud 9, Code Black, Puff, Special K and Voodoo) tested
did not have associated standards and as such whether they would react with ECL
was unknown. As suspected, five of these synthetic compounds did not produce
light when subjected to an applied potential. As the standards that produced light
with chemiluminescence did not react with ECL, it was not surprising that these
compounds also didn't react with ECL given their affinity for Ru(bipy)23+
chemiluminescence. The brand Storm also didn't produce light with ECL, however
this is consistent with the results obtained from the standards PB-22 and 5F-PB-22.
The brands Atomic Bomb and Stoner were not tested as they were known to only
contain PB-22 and 5F-PB-22 in a lesser concentration than Storm.
138
Voodoo (unknown compound) was an exception where light was produced
with both chemiluminescence and ECL. Supanova also produced light with ECL
despite being known to contain PB-22, a synthetic cannabinoid whose standard did
not emit light when subjected to an applied potential. It is proposed that this sample
contains further unidentified components, as other large peaks were visible on the
UV HPLC chromatogram. It is possible that these compounds are unknown to date
as no synthetic cannabinoids could be matched to their respective molecular ions
in mass spectrometry (m/z 216 amu and m/z 375). Concentration of the compounds
and NMR analysis would help, however in this instance insufficient quantities for
NMR analysis of the cannabinoid were able to be collected.
Figure 5.10: Synthetic cannabinoid standard 5F-AKB-48 analogue with ECL detection.
The brand Malibu contains the synthetic cannabinoid 5F-AKB-48, of
which a standard had previously been tested (Figure 5.10) and found to produce
light. The extraction from the herbs produced less light than the standard at 1 mM
due to a lower unknown concentration of extracted synthetic cannabinoid. This
however is important as ECL can still be used to detect the extracted synthetic
cannabinoid and with further optimisation this has the potential to be used in an at-
scene detection test.
‐0.002
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
‐0.5 0 0.5 1 1.5 2
139
Solid Phase Extraction
Six solid phase extraction (SPE) cartridges (C18, C8, SAX, SCX, PCX and
natural silica) were trialled as a sample clean up method to eliminate components
co-extracted from the herbs in order to develop a more targeted identification
protocol. The four organic solvents (acetonitrile, acetone, ethyl acetate and
methanol) identified as the most efficient at extracting the synthetic cannabinoid
from the herb were used as the initial extraction solvent. The solvent extractions
discussed earlier in this chapter indicated that acetonitrile and acetone co-extract
less components from the herbs but still a significant amount of cannabinoid.
Samples were extracted with one of the four organic solvents before being
added to the conditioned cartridge, where it was discovered that the cannabinoid
was breaking through the sorbent on all cartridges, eluting as the sample was added.
However other components were generally retained on the sorbent until washing
with 5% methanol (100% extraction solvent for natural silica cartridge) or elution
with 100% methanol (100% water for natural silica cartridge).
140
A)
B)
Figure 5.11: Chromatograms obtained from each stage of SPE extraction (Plexa PCX cartridge with methanol conditioning and acetonitrile initial extraction) for the brand
“Puff” using UV detection A) actual chromatogram (offset) B) zoomed in chromatogram (offset). Legend; Initial = before SPE, Sample = first pass, Wash = second pass and
Elution = third and final pass.
0
100
200
300
400
500
600
700
800
0 5 10 15 20
Absorban
ce (mAu)
Time (mins)
Initial
Sample
Wash
Elution
0
10
20
30
40
50
60
70
80
90
100
0 5 10 15 20
Absorban
ce (mAu)
Time (mins)
Initial
Sample
Wash
Elution
141
As can be seen from Figure 5.11, the characteristic chromatograms
produced from the extraction of the brand “Puff” have a large peak at
approximately 17 minutes which is the synthetic cannabinoid UR-144. The
chromatograms also show a small peak at the beginning which is likely to be a
solvent front from the extraction solvent. The peak at approximately 13 minutes is
an impurity as it is also present in blank samples. This and all other peaks are very
small in comparison to the cannabinoid peaks. Figure 5.11B shows that there are
still peaks in the sample solution however in comparison to the sample pre-solid
phase extraction it can be noted that they have been reduced.
The solution collected from the wash also contains a large amount of
cannabinoid, which is potentially from the residual sample solution held in the
sorbent as it can’t be allowed to dry out during the SPE process. It is expected that
the cannabinoid should be retained on the sorbent however due to the large
concentration of the cannabinoid, it breaks through the sorbent in the sample. The
synthetic cannabinoid is still present in the elution solution due to the sorbent
retaining a certain amount of the compound. Previous testing indicated that at a
smaller concentration the synthetic cannabinoid is also fully retained until the third
and final pass, however the other non-polar compounds are also retained, leading
to the use of these cartridges as a filtration system with a large concentration of
synthetic cannabinoid [196].
The initial extraction of each sample was analysed with HPLC before SPE
was performed as this provided the opportunity to see how much of the cannabinoid
was lost during the SPE process.
Table 5.11: Results of solid phase extraction for the brand Puff.
Cartridge Conditioning Extraction
solventNo. of Peaks
% Retained
Silica Ethyl acetate Ethyl acetate 4 57
C8 Acetonitrile, water Acetonitrile 3 54
C18 Acetonitrile, water Acetonitrile 3 58
C18 Methanol, water Acetonitrile 2 55
SAX Acetone, water Acetone 5 52
SCX Acetonitrile, water Acetonitrile 5 63
PCX Acetonitrile, water Acetonitrile 4 55
PCX Methanol, water Acetonitrile 4 62
142
As illustrated in Table 5.11, between 52% and 63% of the cannabinoid was
retained in the sample solution. This means that the solid phase extraction process
loses approximately 33%-50% of the cannabinoid, however due to the sensitivity
of the detection this is not an issue. The majority of the cannabinoid was retained
in the sample solution, however the remainder of the cannabinoid was eluted in the
wash and elution stages.
The first pass sample which contains the breakthrough of the synthetic
cannabinoid was analysed using chemiluminescence detection after optimisation of
the method. Although the UV indicated that a few small peaks remained in the
extracted sample after solid phase extraction, when analysed with Ru(bipy)33+ the
only component that produced light was the synthetic cannabinoid. Figure 5.12
illustrates this, as even with a sample where 5 peaks are present in the UV other
than the synthetic cannabinoid peak, the only signal produced by reaction with the
chemiluminescence was the cannabinoid. This illustrates that the methods outlined
above are selective for the synthetic cannabinoids.
Figure 5.12: Chromatogram of an acetonitrile extraction for the brand “Puff” after SPE with a Plexa PCX cartridge with UV (black) and chemiluminescence (red) detection.
0
10
20
30
40
50
60
70
80
90
100
0
50
100
150
200
250
300
350
400
450
0 2 4 6 8 10 12 14 16 18 20
Intensit y (m
Au)
Absorban
ce (mAu)
Time (Mins)
143
Depending on the application and the potential at scene sample matrix
including contamination, different cartridges may best apply and several should be
considered. The C18 cartridge provides the cleanest extraction however only
retains a 55% of the synthetic cannabinoid whereas the Plexa PCX cartridge
produced a larger number of extraneous peaks but retained a larger percentage of
the compound of interest. However, when using selective detection such as
chemiluminescence all tested cartridges can be used as only the synthetic
cannabinoid reacts with the reagent to produce light. For this application the Plexa
PCX cartridge was identified to be the cartridge of choice as it was able to provide
the largest cannabinoid sample.
The work presented in this chapter provides the ground work for the
construction of an at-scene detection test to elucidate if a synthetic cannabinoid is
present on an herbal substrate. The ideal extraction solvent, acetonitrile in
conjunction with a C18 sorbent has been shown to selectively isolate synthetic
cannabinoids in a concentration large enough for detection. Two detection systems
have been trialled here and with further optimisation there is potential for both to
be effective in revealing if these compounds are present. Chemiluminescence and
ECL have both been investigated for use in microfluidics devices and all synthetic
cannabinoids tested here have succeeded with at least one of the two methods.
144
Conclusions
The need for a selective detection method for synthetic cannabinoids has
been highlighted by the vast quantity of available compounds and blends. Polar
organic solvents such as methanol, acetone and acetonitrile provided the most
efficient and selective extraction of the target synthetic cannabinoids from the
herbal materials, which in the products under investigation here were identified as
seven known indole and indazole derivatives. Following solvent extraction, the
synthetic cannabinoids can be determined by chromatographic separation and UV
absorbance detection.
Chemiluminescence with the manganese(IV) or [Ru(bipy)3]3+ reagent was
shown to be a viable alternative for the post-column detection of most synthetic
cannabinoids, although the limits of detection were generally poorer than those
using conventional UV absorbance. The inherent advantages of
chemiluminescence, however, such as superior analytical selectivity and the
relative simplicity of detector components are attractive for the development of
rapid separations and/or at-scene devices for preliminary screening applications.
This preliminary exploration of the chemiluminescence detection of this controlled
class of compound shows that there is considerable scope for further development
in this area.
Solid Phase extraction following a simple solvent extraction with
acetonitrile and a C18 silica cartridge increases the selectivity the extraction and
eliminates the need for a separation technique. This allows the synthetic
cannabinoids to be selectively detected with chemiluminescence without the need
for separation. The methods developed here provide the background chemistry for
an at-scene detection device for synthetic cannabinoids on unknown herbal blends.
145
CHAPTER 6:
TARGETED NMR SPECTROSCOPY FOR SELECTIVE DETERMINATION OF SYNTHETIC CANNABINOIDS
146
Chapter overview
This chapter delves into the use of NMR spectroscopy to directly detect
synthetic cannabinoids on the chemically complex herbal blends. The main focus
of this is on the use of solid state NMR, a novel technique not previously used for
the detection of compounds on already difficult substrates. This technique affords
no damage to the sample which is of particular importance for forensically
significant samples. Experiments investigated included fluorine and both spinning
and static proton NMR. Carbon experiments were also completed with single pulse
excitation and cross polarisation to proton nuclei.
147
Introduction
Synthetic additives sprayed onto plants range from crop protectants such as
herbicides to illicit substances such as synthetic cannabinoids. The emergence of
synthetic cannabinoids has prompted investigations into techniques to detect these
man-made chemicals. The current practices for the determination of synthetic
cannabinoids on herbal substrates involve onerous sample preparation and result in
the destruction of the samples; for example, a solvent or solid phase extraction prior
to chromatographic separation (HPLC or GC) coupled with mass spectral detection
[228-230]. Ideally, multiple techniques could be combined to give an enhanced
fingerprint. While examples exist where solution-state nuclear magnetic resonance
(NMR) spectroscopy has been used for the confirmation of chromatographic peaks
of forensic interest [231, 232] this analysis is performed post-fractionation. Solid-
state NMR is an informative and versatile characterisation technique able to
provide quantitative structural information from inhomogeneous samples in a non-
destructive manner [233-235]. Nevertheless, due to the very weak nuclear
polarisation and correspondingly low inherent signal strength, the ability to directly
detect compounds on surfaces using this method is challenging. The majority of
solid-state NMR studies of surface chemistry involve high surface area materials
such as zeolites [236], metal-organic frameworks [130] or nanoparticles [237].
Although not often used for surface chemistry it is an important technique for
biomolecules [132], nanotechnology [132] and pharmaceuticals [238]. Importantly,
from a forensic perspective where samples are precious, solid-state NMR is a non-
destructive analytical technique that requires little to no sample preparation [239].
This is also particularly important for samples that can degrade when extracted with
solvent. The amount of sample required is generally between 2 mg and 100 mg
depending on the size of the magic angle spinning (MAS) rotor, and typically drug
seizure samples are well in excess of this amount [240].
The technique is not as widely used as solution state NMR due to its low
resolution, however this can be improved with techniques including magic angle
spinning (MAS) and cross polarization (CP) [236]. Magic angle spinning (MAS) is
an essential technique that allows better resolution of peaks as opposed to static
analysis. This involves the rotor being spun at a magic angle with respect to the
148
magnetic field, which narrows the normally wide lines thus improving spectral
resolution [239]. Cross polarisation (CP) is commonly used in solid state NMR to
improve the signal to noise ratio and obtain a more rapid analysis [241]. Cross
polarisation is the transfer of more abundant nuclei such as 1H and 19F to polarise
rare nuclei, including 13C and 15N [242]. Signal enhancement technologies such as
dynamic nuclear polarisation (DNP) have boosted the sensitivity of solid-state
NMR and are opening up a new field of NMR surface characterisation [243].
Previous work utilising solid-state NMR for surface analysis found
sensitivity to be an issue in the detection of surface-immobilised peptides on gold
nanoparticles [244]. In this publication Ashbrook et al. highlight the basic
approaches which can inform on how solids are influenced by anisotropic
(orientation-dependant) interactions including dipolar couplings and chemical shift
anisotrophy which is the driver for the broadening of peaks and subsequent loss in
ability to utlitise the solid state NMR spectra. While magic angle spinning and
decoupling and cross polarisation techniques are used to improve sensitivity and
resolution, solid state NMR has not been used in the capacity highlighted in this
thesis where molecules of interest are interrogated on a solid substrate.
Significant research was also required in the surface analysis of metal
organic frameworks to improve sensitivity where techniques such as the use of
microwaves to induce DNP were required [130]. The techniques outlined here are
standard solid-state NMR methods and equipment, not requiring modification to
the probe as was required in the investigation on surface ethoxy species of acidic
zeolite Y [131]. Early work of solid state NMR has been performed in order to
highlight key features of a surface modification of substrates including silicon [245]
where the reactivity of hydroxyl functional groups on silica with
trimethylchlorosilane (TMCS) was investigated. Interestingly the authors noted
that the technique was able to differentiate between germinal and vicinal silanols
enabling a reactivity mechanism to be postulated. A spectroscopic analysis was
performed however FTIR-PAS and XPS had the capacity to distinguish the features
afforded by this early solid state NMR. This concept was fully realised by Pallister
et al. [246] where quantitative surface coverage analysis was performed on thin
films. This study was able to exploit both 29Si and 13C NMR to generate data on
group V gallium thin films which are wide band gap materials of great interest to
149
optoelectronics and semiconductor manufacturers. In 2000 Ladizhansky and co-
workers [247] interrogated the water binding efficiency on the surface of CdS
nanoparticles that had precipitated from an aqueous solution. In this instance 1H
NMR (MAS) was used to great effect highlighting that three different bindings sites
of the water lead to three different lines in the NMR spectra. Other surfaces have
been studied including nanocrystalline materials such as hydroxyapatite where
Jager et al. [248] were able to fully characterise the material. While solid state NMR
is a fully established technique for the structural characterisation of material on the
atomic scale limited work has been performed on molecules on complex substrates
such as the herbs presented here. Complex backgrounds as demonstrated by these
herbal blends have not been reported before and as such the discovery of the
potential applications outlined here are particularly important.
150
Experimental
Samples
The samples used were brands of herbal based synthetic cannabinoid
products purchased pre ban from various adult and herb stores across Victoria,
Australia. Brands studied were Red Dot, Puff, Bombay Blue, Cloud 9, Special K,
Code Black, Malibu, Atomic Bomb, Storm, Stoner, Supanova and Voodoo.
Damiana (turnera diffusa) was purchased from the happy herb shop, Geelong,
Victoria, Australia. Approximately 100mg of each sample was ground to a fine
powder using mortar and pestle which was then packed into a 4 mm o.d. MAS
NMR rotor for analysis.
Instrumentation and Experiments
A Bruker Avance III 300 MHz Wide Bore was used for all analyses in
conjunction with a 4.0 mm MAS HX probe (Preston, Victoria, Australia). All
experiments were run at 10 kHz MAS spinning rate except for a static 1H
experiment that was carried out on all samples with 4 scans. A 1H MAS experiment
was also carried out with 4 scans. 19F was only carried out on 9 selected samples
with a minimum scan time of 2 hours. 13C CP MAS was carried out at 7.05 T on all
samples with a minimum of 2000 scans and a recycle delay of 5 seconds. 13C single
pulse excitation (SPE) was carried out on 3 samples (Damiana, Bombay Blue and
Cloud 9) at a T1 relaxation time of 10 seconds and on two samples (Damiana and
Bombay Blue) with a T1 relaxation time of 60 seconds.
151
Results and discussion
Given that the cannabinoid is sprayed onto the substrate using a volatile
solvent (typically acetone) it was reasoned that the cannabinoid might be ordered (
potentially crystalline) on the surface and as such discernible from the herbal
substrate. Initial 13C MAS and 1H-13C cross-polarisation (CP) MAS NMR
experiments carried out on the brand Bombay Blue and Damiana confirmed that
the synthetic cannabinoids could be identified using this technique as it was evident
that the peak widths of the synthetic molecule were systematically narrower than
that of the herbal substrate, indicative of an ordered solid phase (Figure 6.1). This
feature facilitated ready identification of the signals arising from the synthetic
molecules. Damiana (turnera diffusa) was used as a reference for all NMR
experiments as this herb is most commonly used as a substrate for these commercial
products. This allowed the background spectrum generated from the substrate to be
subtracted, revealing the detailed synthetic cannabinoid spectrum.
For example, subtracting the 13C spectrum of the substrate herb from that of
the Bombay Blue sample clearly revealed the 13C solid-state NMR spectrum for the
cannabinoid XLR-11 (Table 5.3) and its distinct structural features as shown in
Figure 6.1. The distinct structural features of XLR-11 correspond to the following;
the peaks between 0 and 50 ppm correspond to the aliphatic hydrocarbons, the
peaks between 100 and 150 ppm correspond to the carbons of the aromatic rings,
and the peak at 195 ppm is attributed to the carbonyl group (C=O ketone). The peak
at approximately 80 ppm is consistent with a carbon bound to fluorine.
152
Figure 6.1: Normalised 13C CP MAS solid-state NMR spectra of (top to bottom) damiana herbs, synthetic cannabinoid brand Bombay Blue (containing Synthetic cannabinoid
XLR-11) and Bombay Blue with Damiana subtracted. Spectra were obtained with 16,384 scans.
13C CPMAS experiments were also carried out on the remaining synthetic
cannabinoid products and the spectra allow different cannabinoids to be clearly
differentiated as shown in Figure 6.2 and 6.3. The two synthetic cannabinoids can
be clearly differentiated in Figure 6.2 where the Malibu brand is observed to contain
the synthetic cannabinoid 5F-AKB-48 while the Code Black product contains UR-
144. The correlation of the NMR peak positions (chemical shifts) to the molecular
structure allows certain signals to be used as markers of cannabinoid structure class.
For example, the brand Code Black has a ketone resonance at a chemical shift of
195 ppm as do all samples containing UR-144, XLR-11 and AM-2201, but is absent
in samples that do not contain this feature. In contrast, the brand Malibu has an
amide in its place, resulting in an amide peak at a chemical shift of 160 ppm instead.
Samples that contain PB-22 and 5F-PB-22 such as the brand Storm have an ester
peak at approximately 170 ppm. This differentiation in chemical shift due to
structural differences is illustrated in Figure 6. 3.
200 150 100 50 0
Chemical shift (ppm)
153
Figure 6.2: 13C CPMAS solid-state NMR spectra (normalised and with the Damiana herb signals subtracted) from synthetic cannabinoid brands Malibu (5F-AKB-48) (top)
and Code Black (UR-144) (bottom).
Figure 6.3: 13C CP MAS solid state NMR spectra normalised with the Damiana herb subtracted. Synthetic cannabinoid brands from top to bottom are Malibu, Code Black, Bombay Blue, Puff, Red Dot, Voodoo, Storm, Cloud 9, Stoner, Special K, Atomic Bomb
and Supanova.
200 150 100 50 0
Chemical shift (ppm)
250 200 150 100 50 0 -50
Chemical shift (ppm )
154
As seen in Figure 6.1 in the CP MAS spectra, the signals from the substrate
and the synthetic cannabinoid are comparable in size. Although it indicates the
relative amount of herb to the active molecule, CP MAS experiments are not
generally quantitative. Direct 13C solid-state MAS NMR experiments (i.e. without
cross polarisation to the 1H nuclei) were therefore also carried out, with different
recycle delays. As illustrated in Figure 6.4 it was found that the 13C T1 relaxation
times for the herbal substrate were significantly longer than for the synthetic
coatings. A short recycle delay of 2 seconds therefore effectively suppressed the
signal from the substrate, while a 60 s recycle delay showed both components. This
approach, while somewhat more time consuming due to the weak 13C signal and
long recycle delays required, could potentially allow accurate quantification of the
components.
The concentration of the synthetic cannabinoid also has an impact as even
with 1H-13C cross polarisation, some of the cannabinoid signals known to be present
in low concentrations in the samples from previous analysis, were too weak to
resolve in a reasonable timeframe. As shown in Figure 6.3 the brands Special K,
Atomic Bomb and Supanova lack the peaks corresponding to the synthetic
cannabinoid compound known to be present with only a spectra of the herbs
obtained. These samples are known to contain two cannabinoids but in a lower
concentration than the other samples as was determined in Chapter 5.
155
Figure 6.4: Synthetic cannabinoid brand Bombay Blue with 13C SPE MAS solid state NMR. Red spectra is with T1 relaxation time of 2 seconds and Black spectra is with T1
relaxation time of 60 seconds.
In an effort to circumvent legislation and complicate detection, a large
number of synthetic cannabinoids were modified to include fluorine substituents
often as a fluoropentyl chain [44]. With a gyromagnetic ratio almost four times
larger than 13C and a 100% natural abundance, 19F is both a selective and sensitive
means of compound characterisation. As such, 19F MAS NMR was performed on
brands that were suspected to contain a fluorine (e.g. XLR-11, AM-2201, 5F-PB-
22 and 5F-AKB-48). From these samples a clear signal at a chemical shift of
−218ppm was apparent, characteristic of a primary alkyl fluoride. Synthetic
cannabinoids identified in the brands Puff and Supanova lack fluorine substituents
and no 19F NMR signal could be detected as is supported by the lack of a peak at
−218ppm.
250 200 150 100 50 0 -50
Chemical shift (ppm)
156
Figure 6.5: 19F solid state NMR spectra normalised of synthetic cannabinoid brands (from top to bottom) Malibu (5F-AKB-48), Cloud 9 (UR-144 and XLR-11), Bombay Blue
(XLR-11), Red Dot (AM-2201), Voodoo, Atomic Bomb (5F-PB-22 and PB-22), Code Black (UR-144), Puff (UR-144), Supanova (PB-22) and Damiana (plain herb).
As can be seen in Figure 6.5, a further peak at δ50ppm is present in all
samples including the plain herbal blend, Damiana. This peak was further
investigated to determine if a fluorine was present in the herbal blend by running
an experiment on the rotor and cap without a sample. It was found that this fluorine
peak was also present in the empty rotor spectra and hence is a background signal
possibly from the rotor cap (Figure 6.6). The rotor is known to be made from
zirconium dioxide (ZrO2), but the caps for these rotors can be made from a variety
of different compounds. However, due to the difference in chemical shift it doesn’t
interfere with the fluorine peak from the synthetic cannabinoids and is not an issue
when analysing these samples.
200 100 0 -100 -200 -300 -400
Chemical Shift (ppm)
157
Figure 6.6: Spectra from the empty rotor, plain herb and a fluorine containing synthetic cannabinoid herbal coating.
1H MAS and static 1H NMR (Figure 6.7) were both uninformative on
whether synthetic cannabinoids were present on the herbal blends. It is generally
accepted that 1H solid state NMR is low resolution and provides little information
about the protons. However, information can still be gained from this analysis in
the form of the mobility of the sample. The mobility of the sample relates to how
the structure can reorient depending on the domain. For example with polyethylene
it is in a highly ordered or crystalline domain the polymer chains can only reorient
very slowly [249]. However if they are in non-crystalline domain the polymer
chains are more mobile and as such this can give indication of crystallinity of the
compounds being detected [249]. As the structures and substrates of all he synthetic
cannabinoids are similar the spectra does not significantly vary (Figure 6.7).
158
Figure 6.7: 1H spin solid state NMR normalised spectra of synthetic cannabinoid brands (from top to bottom) Malibu, Supanova, Atomic Bomb, Special K, Stoner, Cloud 9, Storm,
Voodoo, Red Dot, Puff, Bombay Blue and Code Black.
This thesis highlights that sensitive determination of the synthetic
cannabinoid is not an issue. In comparison to GC/MS, the technique currently used
in forensic determination of these compounds, there are some differences and as
such it would be beneficial to use these methods in conjunction with each other. As
previously eluded to solid state NMR requires very little sample preparation in
comparison GC/MS which requires extraction of the compounds into a volatile
solvent. The sample preparation for GC/MS involves grinding the sample into a
powder before adding either methanol or ethanol and ultrasonicating for 10
minutes, centrifuging for a further 5 minutes and then filtering out the herbal blends
[112, 199]. This takes significantly longer than the sample preparation required for
solid state NMR where the sample is only ground to a powder before analysis.
Literature reports the analysis time for GC/MS ranging between 15 minutes and 60
minutes [111, 135, 199]. Whilst this is an improvement on the 3 hour cross
polarisation analysis time it is important to note that a comprehensive study on
improving analysis time was not performed here and could go some way to
enhancing the techniques highlighted in this chapter.
80 60 40 20 0 -20 -40 -60 -80
Chemical shift (ppm)
159
The use of 13C and 19F solid-state NMR experiments have proven to be a
powerful, non-destructive approach for the detection of synthetic cannabinoids as
a component of herbal blends. Using both 13C CP MAS and 19F MAS experiments
key structural features can be readily differentiated. This has implications and is
extremely important for forensic analysis of these as the sample isn’t required to be
destroyed in order to determine if an alterant has been applied. Sample preparation
is straightforward and these experiments are standard and do not require high
magnetic fields or fast spinning rates. This approach will only become more
powerful as sensitivity enhancement techniques such as the recently identified DNP
become more reliable. Beyond the obvious implications for forensic analysis, this
approach has the potential to be used in other applications where herbs or plants
have had compounds added such as herbicides on foreign plant or vegetable
imports.
160
Conclusions
Forensically relevant synthetic cannabinoid compounds on the surface of
herbal substrates have been identified using 13C and 19F solid-state NMR methods.
Further to this, structural differences in compounds can also be determined. Solid-
state NMR is not typically amenable to the study of molecules on surfaces, however
the synthetic cannabinoids studied here provide both high sensitivity and
resolution, indicative of an ordered solid phase. The use of this technique has the
capacity to deliver a selective detection protocol for synthetic cannabinoids on
complex forensic samples in a non-destructive manner. This technique also has the
potential to be used in other applications where a compound has been added to
herbs or plant material.
161
Future Work
This thesis explores chemistry in both pharmaceutical and forensic science
areas and as such the application for future research is limitless. The work presented
here has the potential to be expanded and some ideas have been suggested below.
The mapping techniques delivered in chapters 2, 3 and 4 illustrate the broad
applications to the processing streams undertaken by Sun Pharmaceuticals,
however this technique could be used in other chemical manufacturing companies
to monitor their processes. The exact limits and requirements for these techniques
at each stage of the process need to be developed more precisely to target the exact
information required by the company. This also has the capability to be utilised in
the identification of new impurities introduced in bespoke batches of crop from a
range geographic areas and annual growth cycles to identify issues early in the
process.
Chapter 4 explored the identification of a few select impurities of interest
as this is of significant importance to Sun Pharmaceuticals. The future work in
relation to this portion is to develop techniques to identify the chemical composition
and the potential impact they have on the final product. In particular the impurities
such as those at 285 amu for example are of great importance to identify. It has
been pinpointed from this research where the largest concentration of the 285 amu
impurity is in the process and as such this sample may be explored in order to isolate
enough of this compound for NMR analysis. Other impurities of interest are also
able to be pinpointed as to where they are most concentrated in the stream and as
such also have a greater potential to be isolated for identification. Newly identified
impurities have the potential to be isolated in the process and utilised as future
important medications or as a precursor to make compounds of interest more
straightforward as was the case with the now indispensable compound, thebaine.
The main aim of the synthetic cannabinoids study presented in chapter 5 is
to work towards an at-scene detection test for these herbal substrates. The chemical
background to allow this device to be made has been achieved however placing it
into a device is dependent on engineering as there is a fine line as to what can be
162
achieved with both chemistry and engineering. The future aim of the chemistry
developed here is to fit it into a microfluidic lab on a chip that would allow at scene
detection of these compounds. In theory these chips would be hand held disposable
and cheap devices that could connect to a smart phone or similar to give quick and
reliable results.
Also in chapter 5, Electrochemiluminescence (ECL) was shown to react
with some synthetic cannabinoids, however only the surface of this technique has
been touched. There are a large variety of complexes that can be used in ECL
including different ruthenium and iridium complexes. Whilst work has previously
been done on these complexes, their reactions with the synthetic cannabinoids
could lead to a more targeted approach to these compounds. A fundamental study
of why some of these cannabinoids elicit light with ECL/chemiluminescence and
others don't is also of interest.
In chapter 6, solid state NMR was shown to detect synthetic cannabinoids
on the surface of herbal blends, which may have implications to detect coatings on
other plants. To develop this technique further, a larger sample set of synthetic
cannabinoid standards and different herbal mixtures could be explored to
potentially build a library of different herbal substrates and synthetic cannabinoid
spectra. The non-destructive nature this technique offers has implications in
forensic science and could potentially be used as a screening technique to test if
synthetic cannabinoids are present and if so the potential class of the compounds.
To achieve faster and cleaner results the parameters surrounding these could be
explored and modified to identify the most efficient method. Future work on with
this technique could also include exploring areas such as farming where plants that
chemicals have been applied to could be tested in order to find what remains on the
surface.
Importantly continuation of the work in this thesis is currently underway by
both a PhD and Honours student highlighting the importance of the information
generated in this thesis affords further scientific endeavours. I wish the students the
best of luck in their endeavours as I have greatly enjoyed this study program.
163
REFERENCES
1. Wu, H., J. Guo, S. Chen, X. Liu, Y. Zhou, X. Zhang, and X. Xu, Recent
developments in qualitative and quantitative analysis of phytochemical
constituents and their metabolites using liquid chromatography–mass
spectrometry. Journal of Pharmaceutical and Biomedical Analysis, 2013.
72: p. 267-291.
2. Li, J.W.-H. and J.C. Vederas, Drug Discovery and Natural Products: End
of an Era or an Endless Frontier? Science, 2009. 325(5937): p. 161-165.
3. Vuorela, P., M. Leinonen, P. Saikku, P. Tammela, J.-P. Rauha, T.
Wennberg, and H. Vuorela, Natural products in the process of finding new
drug candidates. Current medicinal chemistry, 2004. 11(11): p. 1375-1389.
4. Verkouteren, J.R. and J.L. Staymates, Reliability of ion mobility
spectrometry for qualitative analysis of complex, multicomponent illicit
drug samples. Forensic Science International, 2011. 206(1–3): p. 190-196.
5. Truta, L., A.L. Castro, S. Tarelho, P. Costa, M.G.F. Sales, and H.M.
Teixeira, Antidepressants detection and quantification in whole blood
samples by GC–MS/MS, for forensic purposes. Journal of Pharmaceutical
and Biomedical Analysis, 2016. 128: p. 496-503.
6. Drugs, Poisons and Controlled Substances Act 1981, in 9719. 2017:
Australia. p. 398.
7. Schiff Jr, P.L., Opium and its alkaloids. American Journal of
Pharmaceutical Education, 2002. 66(2): p. 186.
8. Askitopoulou, H., I.A. Ramoutsaki, and E. Konsolaki, Archaeological
evidence on the use of opium in the Minoan world. International Congress
Series, 2002. 1242: p. 23-29.
9. Berridge, V.a.E., G., Opium and the people. 1 ed. 1981, Great Britain: Allen
lane/St Martin's press. 370.
10. Brownstein, M.J., A brief history of opiates, opioid peptides, and opioid
receptors. Proceedings of the National Academy of Sciences, 1993. 90(12):
p. 5391-5393.
164
11. Leonard, E., W. Runguphan, S. O'Connor, and K.J. Prather, Opportunities
in metabolic engineering to facilitate scalable alkaloid production. Nature
chemical biology, 2009. 5(5): p. 292-300.
12. Nyman, U.L.F., Selection for high thebaine/low morphine content (cpv.
Morph: The) in Papaver somniferum L. Hereditas, 1978. 89(1): p. 43-50.
13. Nyman, U. and O. Hall, Some varieties of Papaver somniferum L. with
changed morphinane alkaloid content. Hereditas, 1976. 84(1): p. 69-76.
14. Chaterjee, A., S. Shukla, P. Mishra, A. Rastogi, and S.P. Singh, Prospects
of in vitro production of thebaine in opium poppy (Papaver somniferum L.).
Industrial Crops and Products, 2010. 32(3): p. 668-670.
15. Chatterjee, A., S. Shukla, B. Kishore Mishra, A. Rastogi, and S. Prasad
Singh, Induction of variability through mutagenesis in opium poppy
(Papaver somniferum L.). Turkish Journal of Agriculture & Forestry, 2012.
36(1): p. 1-11.
16. Alagoz, Y., T. Gurkok, B. Zhang, and T. Unver, Manipulating the
Biosynthesis of Bioactive Compound Alkaloids for Next-Generation
Metabolic Engineering in Opium Poppy Using CRISPR-Cas 9 Genome
Editing Technology. Scientific Reports, 2016. 6: p. 30910.
17. Acevska, J., A. Dimitrovska, G. Stefkov, K. Brezovska, M. Karapandzova,
and S. Kulevanova, Development and validation of a reversed-Phase HPLC
method for determination of alkaloids from Papaver somniferum
L.(Papaveraceae). Journal of AOAC International, 2012. 95(2): p. 399-405.
18. Walwyn, W.M., K.A. Miotto, and C.J. Evans, Opioid pharmaceuticals and
addiction: The issues, and research directions seeking solutions. Drug and
Alcohol Dependence, 2010. 108(3): p. 156-165.
19. Yoshimatsu, K., F. Kiuchi, K. Shimomura, and Y. Makino, A rapid and
reliable solid-phase extraction method for high-performance liquid
chromatographic analysis of opium alkaloids from Papaver plants.
Chemical and pharmaceutical bulletin, 2005. 53(11): p. 1446-1450.
20. Gossop, M. and F. Keaney, Research Note—Prescribing Diamorphine for
Medical Conditions: A Very British Practice. Journal of Drug Issues, 2004.
34(2): p. 441-450.
165
21. Anthony John Fist, C.J.B., Wayne Lyle Gerlach,, PRODUCTION OF
THEBAINE AND ORIPAVINE. 2004, Tasmanian Alkaloids PTY. Ltd.,
Westbury (AU).
22. Gómez-Serranillos, M.P., O.M. Palomino, E. Carretero, and A. Villar,
Analytical study and analgesic activity of oripavine from Papaver
somniferum L. Phytotherapy Research, 1998. 12(5): p. 346-349.
23. Den Bosch, J.J.K.-V., C.A. Salemink, and I. Khan, Biological activity of
the alkaloids of Papaver bracteatum lindl. Journal of Ethnopharmacology,
1981. 3(1): p. 21-38.
24. Duke, J.A., Utilization of Papaver. Economic Botany, 1973. 27(4): p. 390-
400.
25. Barber, R.B. and H. Rapoport, Synthesis of thebaine and oripavine from
codeine and morphine. Journal of medicinal chemistry, 1975. 18(11): p.
1074-1077.
26. GlaxoSmithKline. GSK GlaxoSmithKline. 2011 7/5/2014]; Available from:
http://www.gsk.com.au/default.aspx.
27. Rosenbaum, C.D., S.P. Carreiro, and K.M. Babu, Here Today, Gone
Tomorrow…and Back Again? A Review of Herbal Marijuana Alternatives
(K2, Spice), Synthetic Cathinones (Bath Salts), Kratom, Salvia divinorum,
Methoxetamine, and Piperazines. Journal of Medical Toxicology, 2012.
8(1): p. 15-32.
28. Hudson, S. and J. Ramsey, The emergence and analysis of synthetic
cannabinoids. Drug Testing and Analysis, 2011. 3(7-8): p. 466-478.
29. Vardakou, I., C. Pistos, and C. Spiliopoulou, Spice drugs as a new trend:
Mode of action, identification and legislation. Toxicology Letters, 2010.
197(3): p. 157-162.
30. Glass, M. and J.K. Northup, Agonist Selective Regulation of G Proteins by
Cannabinoid CB<sub>1</sub> and CB<sub>2</sub> Receptors.
Molecular Pharmacology, 1999. 56(6): p. 1362-1369.
31. Huffman, J.W., D. Dai, B.R. Martin, and D.R. Compton, Design, Synthesis
and Pharmacology of Cannabimimetic Indoles. Bioorganic & Medicinal
Chemistry Letters, 1994. 4(4): p. 563-566.
32. Denooz, R., J.-C. Vanheugen, M. Frederich, P. de Tullio, and C. Charlier,
Identification and Structural Elucidation of Four Cannabimimetic
166
Compounds (RCS-4, AM-2201, JWH-203 and JWH-210) in Seized
Products. Journal of Analytical Toxicology, 2013. 37(2): p. 56-63.
33. Coppola, M., R. Mondola, F. Oliva, R.L. Picci, D. Ascheri, and F. Trivelli,
Chapter 63 - Treating the Phenomenon of New Psychoactive Substances:
Synthetic Cannabinoids and Synthetic Cathinones A2 - Preedy, Victor R, in
Neuropathology of Drug Addictions and Substance Misuse. 2016,
Academic Press: San Diego. p. 679-686.
34. Lindigkeit, R., A. Boehme, I. Eiserloh, M. Luebbecke, M. Wiggermann, L.
Ernst, and T. Beuerle, Spice: A never ending story? Forensic Science
International, 2009. 191(1–3): p. 58-63.
35. Coulter, C., M. Garnier, and C. Moore, Synthetic Cannabinoids in Oral
Fluid. Journal of Analytical Toxicology, 2011. 35(7): p. 424-430.
36. Park, Y., C. Lee, H. Lee, J. Pyo, J. Jo, J. Lee, H. Choi, S. Kim, R.S. Hong,
Y. Park, B.Y. Hwang, S. Choe, and J.H. Jung, Identification of a new
synthetic cannabinoid in a herbal mixture: 1-butyl-3-(2-
methoxybenzoyl)indole. Forensic Toxicology, 2013. 31(2): p. 187-196.
37. Ciolino, L.A., Quantitation of Synthetic Cannabinoids in Plant Materials
Using High Performance Liquid Chromatography with UV Detection
(Validated Method). Journal of Forensic Sciences, 2015. 60(5): p. 1171-
1181.
38. Ogata, J., N. Uchiyama, R. Kikura-Hanajiri, and Y. Goda, DNA sequence
analyses of blended herbal products including synthetic cannabinoids as
designer drugs. Forensic Science International, 2013. 227(1–3): p. 33-41.
39. Luke, A.D. and C.V. Joris, Cannabis Concerns: Increased Potency,
Availability and Synthetic Analogues. Current Drug Abuse Reviews, 2014.
7(2): p. 67-68.
40. Müller, H.H., J. Kornhuber, and W. Sperling, The behavioral profile of
spice and synthetic cannabinoids in humans. Brain Research Bulletin, 2016.
126, Part 1: p. 3-7.
41. Squires, M., Geelong 'fake drugs' crisis, in Geelong Advertiser. 2013.
42. Castaneto, M.S., D.A. Gorelick, N.A. Desrosiers, R.L. Hartman, S. Pirard,
and M.A. Huestis, Synthetic cannabinoids: Epidemiology,
pharmacodynamics, and clinical implications. Drug and Alcohol
Dependence, 2014. 144: p. 12-41.
167
43. Maxwell, P., Synthetic cannabinomimetics and ‘legal highs’. Journal of
Pharmacy Practice and Research, 2014. 44(4): p. 238-239.
44. Banister, S.D., J. Stuart, R.C. Kevin, A. Edington, M. Longworth, S.M.
Wilkinson, C. Beinat, A.S. Buchanan, D.E. Hibbs, M. Glass, M. Connor,
I.S. McGregor, and M. Kassiou, Effects of Bioisosteric Fluorine in
Synthetic Cannabinoid Designer Drugs JWH-018, AM-2201, UR-144,
XLR-11, PB-22, 5F-PB-22, APICA, and STS-135. ACS Chemical
Neuroscience, 2015. 6(8): p. 1445-1458.
45. Banister, S.D., M. Moir, J. Stuart, R.C. Kevin, K.E. Wood, M. Longworth,
S.M. Wilkinson, C. Beinat, A.S. Buchanan, M. Glass, M. Connor, I.S.
McGregor, and M. Kassiou, Pharmacology of Indole and Indazole
Synthetic Cannabinoid Designer Drugs AB-FUBINACA, ADB-
FUBINACA, AB-PINACA, ADB-PINACA, 5F-AB-PINACA, 5F-ADB-
PINACA, ADBICA, and 5F-ADBICA. ACS Chemical Neuroscience, 2015.
6(9): p. 1546-1559.
46. Harris, C.R. and A. Brown, Synthetic Cannabinoid Intoxication: A Case
Series and Review. The Journal of Emergency Medicine, 2013. 44(2): p.
360-366.
47. Hopkins, C.Y. and B.L. Gilchrist, A case of cannabinoid hyperemesis
syndrome caused by synthetic cannabinoids. The Journal of emergency
medicine, 2013. 45(4): p. 544-546.
48. Gerostamoulos, D., O.H. Drummer, and N.W. Woodford, Deaths linked to
synthetic cannabinoids. Forensic science, medicine, and pathology, 2015.
11(3): p. 478-478.
49. Adamowicz, P., Fatal intoxication with synthetic cannabinoid MDMB-
CHMICA. Forensic Science International, 2016. 261: p. e5-e10.
50. (AIHW), A.I.o.H.a.W. National Drug Strategy Household Survey
(NDSHS). 2013 2017 [cited 2017; Available from:
http://www.aihw.gov.au/alcohol-and-other-drugs/data-sources/ndshs-
2013/.
51. Adam R Winstock, Monica Barratt, Jason Ferris, and Larissa J. Maier.
Global Drug Survey. 2014 [cited 2017; Available from:
https://www.globaldrugsurvey.com/past-findings/.
168
52. Government, V.S. health.vic Synthetic cannabinoids. 2017 [cited 2017;
Available from: https://www2.health.vic.gov.au/public-health/drugs-and-
poisons/drugs-poisons-legislation/synthetic-cannabinoids.
53. Andrews, D., Synthetic drugs banned and tough new ice laws L.
Maksimovic, Editor. 2017, Premier of Victoria.
54. Znaleziona, J., P. Ginterová, J. Petr, P. Ondra, I. Válka, J. Ševčík, J.
Chrastina, and V. Maier, Determination and identification of synthetic
cannabinoids and their metabolites in different matrices by modern
analytical techniques – a review. Analytica Chimica Acta, 2015. 874: p. 11-
25.
55. Francis, P.S., J.L. Adcock, J.W. Costin, S.D. Purcell, F.M. Pfeffer, and
N.W. Barnett, Chemiluminescence detection of opium poppy (Papaver
somniferum) alkaloids. Journal of pharmaceutical and biomedical analysis,
2008. 48(3): p. 508-518.
56. Siddiqui, M.R., Z.A. AlOthman, and N. Rahman, Analytical techniques in
pharmaceutical analysis: A review. Arabian Journal of Chemistry, 2017.
10, Supplement 1: p. S1409-S1421.
57. Altria, K.D. and D. Elder, Overview of the status and applications of
capillary electrophoresis to the analysis of small molecules. Journal of
Chromatography A, 2004. 1023(1): p. 1-14.
58. Gika, H., G. Kaklamanos, P. Manesiotis, and G. Theodoridis,
Chromatography: High-Performance Liquid Chromatography, in
Encyclopedia of Food and Health. 2016, Academic Press: Oxford. p. 93-
99.
59. Houck, M.M. and J.A. Siegel, Fundamentals of forensic science. 2009:
Academic Press.
60. Hanai, T., HPLC: a practical guide. Vol. 6. 1999: Royal Society of
Chemistry.
61. Kavanagh, P., A. Grigoryev, S. Savchuk, I. Mikhura, and A. Formanovsky,
UR-144 in products sold via the Internet: Identification of related
compounds and characterization of pyrolysis products. Drug Testing and
Analysis, 2013. 5(8): p. 683-692.
62. Mostafavi, A., G. Abedi, A. Jamshidi, D. Afzali, and M. Talebi,
Development and validation of a HPLC method for the determination of
169
buprenorphine hydrochloride, naloxone hydrochloride and
noroxymorphone in a tablet formulation. Talanta, 2009. 77(4): p. 1415-
1419.
63. Costin, J.W., S.W. Lewis, S.D. Purcell, L.R. Waddell, P.S. Francis, and
N.W. Barnett, Rapid determination of Papaver somniferum alkaloids in
process streams using monolithic column high-performance liquid
chromatography with chemiluminescence detection. Analytica Chimica
Acta, 2007. 597(1): p. 19-23.
64. Stoll, D.R., X. Li, X. Wang, P.W. Carr, S.E.G. Porter, and S.C. Rutan, Fast,
comprehensive two-dimensional liquid chromatography. Journal of
Chromatography A, 2007. 1168(1–2): p. 3-43.
65. Stevenson, P.G., M. Mnatsakanyan, G. Guiochon, and R.A. Shalliker, Peak
picking and the assessment of separation performance in two-dimensional
high performance liquid chromatography. Analyst, 2010. 135(7): p. 1541-
1550.
66. Bassanese, D.N., B.J. Holland, X.A. Conlan, P.S. Francis, N.W. Barnett,
and P.G. Stevenson, Protocols for finding the most orthogonal dimensions
for two-dimensional high performance liquid chromatography. Talanta,
2015. 134: p. 402-408.
67. Fairchild, J.N., K. Horváth, and G. Guiochon, Approaches to
comprehensive multidimensional liquid chromatography systems. Journal
of Chromatography A, 2009. 1216(9): p. 1363-1371.
68. Stevenson, P.G., D.N. Bassanese, N.W. Barnett, and X.A. Conlan,
Improved 2D-HPLC of red wine by incorporating pre-process signal-
smoothing algorithms. Journal of Separation Science, 2013. 36(21-22): p.
3503-3510.
69. Furton, K.G., M. Kusano, and M.S. Macias, GAS CHROMATOGRAPHY
| Forensic Applications A2 - Wilson, Ian D, in Encyclopedia of Separation
Science. 2007, Academic Press: Oxford. p. 1-8.
70. Mohr, S., J.A. Weiß, J. Spreitz, and M.G. Schmid, Chiral separation of new
cathinone- and amphetamine-related designer drugs by gas
chromatography–mass spectrometry using trifluoroacetyl-l-prolyl chloride
as chiral derivatization reagent. Journal of Chromatography A, 2012. 1269:
p. 352-359.
170
71. Singh, S., T. Handa, M. Narayanam, A. Sahu, M. Junwal, and R.P. Shah, A
critical review on the use of modern sophisticated hyphenated tools in the
characterization of impurities and degradation products. Journal of
Pharmaceutical and Biomedical Analysis, 2012. 69: p. 148-173.
72. Elder, D.P., D. Snodin, and A. Teasdale, Control and analysis of hydrazine,
hydrazides and hydrazones—Genotoxic impurities in active pharmaceutical
ingredients (APIs) and drug products. Journal of Pharmaceutical and
Biomedical Analysis, 2011. 54(5): p. 900-910.
73. Płonka, M., S. Walorczyk, and M. Miszczyk, Chromatographic methods for
the determination of active substances and characterization of their
impurities in pesticide formulations. TrAC Trends in Analytical Chemistry,
2016. 85, Part B: p. 67-80.
74. Eckre, D., L. Fadness, G. Rankin, and C. Bridge, Thermal Degradation of
Synthetic Cannabinoids Containing a Cyclopropyl Group.
75. Denis, C.M. and N.E. Baryla, Determination of piperazine in
pharmaceutical drug substances using capillary electrophoresis with
indirect UV detection. Journal of Chromatography A, 2006. 1110(1–2): p.
268-271.
76. Cruces-Blanco, C., L. Gámiz–Gracia, and A.M. García-Campaña,
Applications of capillary electrophoresis in forensic analytical chemistry.
TrAC Trends in Analytical Chemistry, 2007. 26(3): p. 215-226.
77. Hindson, B.J., P.S. Francis, S.D. Purcell, and N.W. Barnett, Determination
of opiate alkaloids in process liquors using capillary electrophoresis.
Journal of Pharmaceutical and Biomedical Analysis, 2007. 43(3): p. 1164-
1168.
78. Gerardi, R.D., N.W. Barnett, and S.W. Lewis, Analytical applications of
tris(2,2′-bipyridyl)ruthenium(III) as a chemiluminescent reagent. Analytica
Chimica Acta, 1999. 378(1–3): p. 1-41.
79. Adcock, J.L., C.J. Barrow, N.W. Barnett, X.A. Conlan, C.F. Hogan, and
P.S. Francis, Chemiluminescence and electrochemiluminescence detection
of controlled drugs. Drug Testing and Analysis, 2011. 3(3): p. 145-160.
80. Percy, D.W., J.L. Adcock, X.A. Conlan, N.W. Barnett, M.E. Gange, L.K.
Noonan, L.C. Henderson, and P.S. Francis, Determination of Citrus
171
aurantium protoalkaloids using HPLC with acidic potassium permanganate
chemiluminescence detection. Talanta, 2010. 80(5): p. 2191-2195.
81. McDermott, G.P., X.A. Conlan, L.K. Noonan, J.W. Costin, M.
Mnatsakanyan, R.A. Shalliker, N.W. Barnett, and P.S. Francis, Screening
for antioxidants in complex matrices using high performance liquid
chromatography with acidic potassium permanganate chemiluminescence
detection. Analytica Chimica Acta, 2011. 684(1–2): p. 134-141.
82. Khataee, A., R. Lotfi, A. Hasanzadeh, M. Iranifam, and S.W. Joo, Flow-
injection chemiluminescence analysis for sensitive determination of
atenolol using cadmium sulfide quantum dots. Spectrochimica Acta Part A:
Molecular and Biomolecular Spectroscopy, 2016. 157: p. 88-95.
83. Nalewajko-Sieliwoniuk, E., J. Malejko, M. Święczkowska, and A.
Kowalewska, A study on the selection of chemiluminescence system for the
flow injection determination of the total polyphenol index of plant-derived
foods. Food Chemistry, 2015. 176: p. 175-183.
84. Francis, P.S., J.W. Costin, X.A. Conlan, S.A. Bellomarino, J.A. Barnett,
and N.W. Barnett, A rapid antioxidant assay based on acidic potassium
permanganate chemiluminescence. Food Chemistry, 2010. 122(3): p. 926-
929.
85. Lenehan, C.E., N.W. Barnett, S.W. Lewis, and K.M. Essery, Preliminary
evaluation of dual acidic potassium permanganate and tris (2, 2′-bipyridyl)
ruthenium (II) chemiluminescence detection for the HPLC determination of
Papaver somniferum alkaloids. Australian journal of chemistry, 2004.
57(10): p. 1001-1004.
86. Brown, A.J., C.E. Lenehan, P.S. Francis, D.E. Dunstan, and N.W. Barnett,
Soluble manganese(IV) as a chemiluminescence reagent for the
determination of opiate alkaloids, indoles and analytes of forensic interest.
Talanta, 2007. 71(5): p. 1951-1957.
87. Li, Y.H. and J.R. Lü, Determination of three fluoroquinolone drugs by flow
injection chemiluminescence. Fenxi Huaxue/ Chinese Journal of Analytical
Chemistry, 2007. 35(5): p. 743-746.
88. Holland, B.J., P.S. Francis, B. Li, T. Tsuzuki, J.L. Adcock, N.W. Barnett,
and X.A. Conlan, Chemiluminescence detection of cannabinoids and
172
related compounds with acidic potassium permanganate. Drug Testing and
Analysis, 2012. 4(7-8): p. 675-679.
89. Camenzuli, M., J.M. Terry, R.A. Shalliker, X.A. Conlan, N.W. Barnett, and
P.S. Francis, Parallel segmented outlet flow high performance liquid
chromatography with multiplexed detection. Analytica Chimica Acta, 2013.
803: p. 154-159.
90. Theakstone, A.G., Z.M. Smith, J.M. Terry, N.W. Barnett, and P.S. Francis,
Chemiluminescence detection of MDMA in street drug samples using
tris(2,2′-bipyridine)ruthenium(III). Drug Testing and Analysis, 2015. 7(5):
p. 428-432.
91. McDermott, G.P., P. Jones, N.W. Barnett, D.N. Donaldson, and P.S.
Francis, Stable tris (2, 2′-bipyridine) ruthenium (III) for chemiluminescence
detection. Analytical chemistry, 2011. 83(13): p. 5453-5457.
92. Gorman, B.A., P.S. Francis, and N.W. Barnett, Tris(2,2'-
bipyridyl)ruthenium(II) chemiluminescence. Analyst, 2006. 131(5): p. 616-
639.
93. Waite, R.J., G.J. Barbante, N.W. Barnett, E.M. Zammit, and P.S. Francis,
Chemiluminescence detection of piperazine designer drugs and related
compounds using tris(2,2′-bipyridine)ruthenium(III). Talanta, 2013. 116: p.
1067-1072.
94. Adcock, J.L., P.S. Francis, and N.W. Barnett, Acidic potassium
permanganate as a chemiluminescence reagent—A review. Analytica
chimica acta, 2007. 601(1): p. 36-67.
95. Hindson, B.J. and N.W. Barnett, Analytical applications of acidic potassium
permanganate as a chemiluminescence reagent. Analytica chimica acta,
2001. 445(1): p. 1-19.
96. Adcock, J.L., N.W. Barnett, C.J. Barrow, and P.S. Francis, Advances in the
use of acidic potassium permanganate as a chemiluminescence reagent: a
review. Analytica chimica acta, 2014. 807: p. 9-28.
97. Adcock, J.L., Z.M. Smith, N.W. Barnett, G.J. Barbante, E.H. Doeven, and
P.S. Francis, A review of recent advances in chemiluminescence detection
using nano-colloidal manganese (IV). Analytica chimica acta, 2014. 848:
p. 1-9.
173
98. Brown, A.J., P.S. Francis, J.L. Adcock, K.F. Lim, and N.W. Barnett,
Manganese(III) and manganese(IV) as chemiluminescence reagents: A
review. Analytica Chimica Acta, 2008. 624(2): p. 175-183.
99. Terry, J.M., Z.M. Smith, G.P. McDermott, R.J. Waite, N.W. Barnett, L.C.
Henderson, J.M. Altimari, and P.S. Francis, Chemiluminescence detection
of amino acids and related compounds using acidic potassium
permanganate, manganese(IV) or tris(2,2′-bipyridine)ruthenium(III).
Talanta, 2012. 99: p. 1051-1056.
100. Nalewajko-Sieliwoniuk, E., I. Tarasewicz, and A. Kojło, Flow injection
chemiluminescence determination of the total phenolics levels in plant-
derived beverages using soluble manganese(IV). Analytica Chimica Acta,
2010. 668(1): p. 19-25.
101. Malejko, J., E. Nalewajko-Sieliwoniuk, J. Nazaruk, J. Siniło, and A. Kojło,
Determination of the total polyphenolic content in Cirsium palustre (L.)
leaves extracts with manganese(IV) chemiluminescence detection. Food
Chemistry, 2014. 152: p. 155-161.
102. Richter, M.M., Electrochemiluminescence (ECL). Chemical Reviews,
2004. 104(6): p. 3003-3036.
103. Tokel, N.E. and A.J. Bard, Electrogenerated chemiluminescence. IX.
Electrochemistry and emission from systems containing tris (2, 2′-
bipyridine) ruthenium (II) dichloride. J. Am. Chem. Soc, 1972. 94(8): p.
2862-2863.
104. Kirschbaum, S.E.K. and A.J. Baeumner, A review of
electrochemiluminescence (ECL) in and for microfluidic analytical devices.
Analytical and Bioanalytical Chemistry, 2015. 407(14): p. 3911-3926.
105. Miao, W., Electrogenerated Chemiluminescence and Its Biorelated
Applications. Chemical Reviews, 2008. 108(7): p. 2506-2553.
106. Hosseini, M., M.R.K. Pur, P. Norouzi, M.R. Moghaddam, and M.R.
Ganjali, An enhanced electrochemiluminescence sensor modified with a
Ru(bpy)32 +/Yb2O3 nanoparticle/nafion composite for the analysis of
methadone samples. Materials Science and Engineering: C, 2017. 76: p.
483-489.
107. McMurry, J., Organic chemistry. 2008: Belmont, CA : Thomson-
Brooks/Cole, c2008.7th ed.
174
108. Blackman, A., S.E. Bottle, S. Schmid, M. Mocerino, and U. Wille,
CHEMISTRY 3E. 2015, Melbourne, AUSTRALIA: Wiley.
109. Auwärter, V., S. Dresen, W. Weinmann, M. Müller, M. Pütz, and N.
Ferreirós, ‘Spice’ and other herbal blends: harmless incense or cannabinoid
designer drugs? Journal of Mass Spectrometry, 2009. 44(5): p. 832-837.
110. Ernst, L., K. Krüger, R. Lindigkeit, H.-M. Schiebel, and T. Beuerle,
Synthetic cannabinoids in “spice-like” herbal blends: First appearance of
JWH-307 and recurrence of JWH-018 on the German market. Forensic
Science International, 2012. 222(1–3): p. 216-222.
111. Uchiyama, N., M. Kawamura, R. Kikura-Hanajiri, and Y. Goda,
Identification of two new-type synthetic cannabinoids, N-(1-adamantyl)-1-
pentyl-1H-indole-3-carboxamide (APICA) and N-(1-adamantyl)-1-pentyl-
1H-indazole-3-carboxamide (APINACA), and detection of five synthetic
cannabinoids, AM-1220, AM-2233, AM-1241, CB-13 (CRA-13), and AM-
1248, as designer drugs in illegal products. Forensic Toxicology, 2012.
30(2): p. 114-125.
112. Uchiyama, N., R. Kikura-Hanajiri, J. Ogata, and Y. Goda, Chemical
analysis of synthetic cannabinoids as designer drugs in herbal products.
Forensic Science International, 2010. 198(1–3): p. 31-38.
113. Ammann, J., J.M. McLaren, D. Gerostamoulos, and J. Beyer, Detection and
Quantification of New Designer Drugs in Human Blood: Part 1 – Synthetic
Cannabinoids. Journal of Analytical Toxicology, 2012. 36(6): p. 372-380.
114. Dresen, S., S. Kneisel, W. Weinmann, R. Zimmermann, and V. Auwärter,
Development and validation of a liquid chromatography–tandem mass
spectrometry method for the quantitation of synthetic cannabinoids of the
aminoalkylindole type and methanandamide in serum and its application to
forensic samples. Journal of Mass Spectrometry, 2011. 46(2): p. 163-171.
115. Dziadosz, M., J.-P. Weller, M. Klintschar, and J. Teske, Scheduled multiple
reaction monitoring algorithm as a way to analyse new designer drugs
combined with synthetic cannabinoids in human serum with liquid
chromatography–tandem mass spectrometry. Journal of Chromatography
B, 2013. 929: p. 84-89.
116. Hutter, M., S. Kneisel, V. Auwärter, and M.A. Neukamm, Determination
of 22 synthetic cannabinoids in human hair by liquid chromatography–
175
tandem mass spectrometry. Journal of Chromatography B, 2012. 903: p.
95-101.
117. Salomone, A., E. Gerace, F. D'Urso, D. Di Corcia, and M. Vincenti,
Simultaneous analysis of several synthetic cannabinoids, THC, CBD and
CBN, in hair by ultra-high performance liquid chromatography tandem
mass spectrometry. Method validation and application to real samples.
Journal of Mass Spectrometry, 2012. 47(5): p. 604-610.
118. Kavanagh, P., A. Grigoryev, A. Melnik, S. Savchuk, A. Simonov, and V.
Rozhanets, Detection and tentative identification of urinary phase I
metabolites of phenylacetylindole cannabimimetics JWH-203 and JWH-
251, by GC–MS and LC–MS/MS. Journal of Chromatography B, 2013.
934: p. 102-108.
119. Grigoryev, A., A. Melnik, S. Savchuk, A. Simonov, and V. Rozhanets, Gas
and liquid chromatography–mass spectrometry studies on the metabolism
of the synthetic phenylacetylindole cannabimimetic JWH-250, the
psychoactive component of smoking mixtures. Journal of Chromatography
B, 2011. 879(25): p. 2519-2526.
120. Antelo-Domínguez, Á., J. Ángel Cocho, M. Jesús Tabernero, A. María
Bermejo, P. Bermejo-Barrera, and A. Moreda-Piñeiro, Simultaneous
determination of cocaine and opiates in dried blood spots by electrospray
ionization tandem mass spectrometry. Talanta, 2013. 117: p. 235-241.
121. Rana, S., R.K. Garg, and A. Singla, Rapid analysis of urinary opiates using
fast gas chromatography–mass spectrometry and hydrogen as a carrier gas.
Egyptian Journal of Forensic Sciences, 2014. 4(3): p. 100-107.
122. Imbert, L., S. Dulaurent, M. Mercerolle, J. Morichon, G. Lachâtre, and J.M.
Gaulier, Development and validation of a single LC–MS/MS assay
following SPE for simultaneous hair analysis of amphetamines, opiates,
cocaine and metabolites. Forensic Science International, 2014. 234: p. 132-
138.
123. Blair, B., A. Nikolaus, C. Hedman, R. Klaper, and T. Grundl, Evaluating
the degradation, sorption, and negative mass balances of pharmaceuticals
and personal care products during wastewater treatment. Chemosphere,
2015. 134: p. 395-401.
176
124. Archer, E., B. Petrie, B. Kasprzyk-Hordern, and G.M. Wolfaardt, The fate
of pharmaceuticals and personal care products (PPCPs), endocrine
disrupting contaminants (EDCs), metabolites and illicit drugs in a WWTW
and environmental waters. Chemosphere, 2017. 174: p. 437-446.
125. Boleda, M.R., M.T. Galceran, and F. Ventura, Trace determination of
cannabinoids and opiates in wastewater and surface waters by ultra-
performance liquid chromatography–tandem mass spectrometry. Journal of
Chromatography A, 2007. 1175(1): p. 38-48.
126. Ziegler, J., P.J. Facchini, R. Geißler, J. Schmidt, C. Ammer, R. Kramell, S.
Voigtländer, A. Gesell, S. Pienkny, and W. Brandt, Evolution of morphine
biosynthesis in opium poppy. Phytochemistry, 2009. 70(15–16): p. 1696-
1707.
127. Furuya, T., A. Ikuta, and K. Syōno, Alkaloids from callus tissue of Papaver
somniferum. Phytochemistry, 1972. 11(10): p. 3041-3044.
128. Stranska, I., M. Skalicky, J. Novak, E. Matyasova, and V. Hejnak, Analysis
of selected poppy (Papaver somniferum L.) cultivars: Pharmaceutically
important alkaloids. Industrial Crops and Products, 2013. 41: p. 120-126.
129. Macomber, R.S., A complete introduction to modern NMR spectroscopy.
1998: Wiley New York.
130. Rossini, A.J., A. Zagdoun, M. Lelli, J. Canivet, S. Aguado, O. Ouari, P.
Tordo, M. Rosay, W.E. Maas, C. Copéret, D. Farrusseng, L. Emsley, and
A. Lesage, Dynamic Nuclear Polarization Enhanced Solid-State NMR
Spectroscopy of Functionalized Metal–Organic Frameworks. Angewandte
Chemie, 2012. 124(1): p. 127-131.
131. Wang, W., J. Jiao, Y. Jiang, S.S. Ray, and M. Hunger, Formation and
Decomposition of Surface Ethoxy Species on Acidic Zeolite Y.
ChemPhysChem, 2005. 6(8): p. 1467-1469.
132. Dybowski*, C. and S. Bai, Solid-state NMR spectroscopy. Analytical
chemistry, 2008. 80(12): p. 4295-4300.
133. Uchiyama, N., M. Kawamura, R. Kikura-Hanajiri, and Y. Goda,
Identification and quantitation of two cannabimimetic phenylacetylindoles
JWH-251 and JWH-250, and four cannabimimetic naphthoylindoles JWH-
081, JWH-015, JWH-200, and JWH-073 as designer drugs in illegal
products. Forensic Toxicology, 2011. 29(1): p. 25-37.
177
134. Simolka, K., R. Lindigkeit, H.-M. Schiebel, U. Papke, L. Ernst, and T.
Beuerle, Analysis of synthetic cannabinoids in “spice-like” herbal highs:
snapshot of the German market in summer 2011. Analytical and
Bioanalytical Chemistry, 2012. 404(1): p. 157-171.
135. Moosmann, B., S. Kneisel, U. Girreser, V. Brecht, F. Westphal, and V.
Auwärter, Separation and structural characterization of the synthetic
cannabinoids JWH-412 and 1-[(5-fluoropentyl)-1H-indol-3yl]-(4-
methylnaphthalen-1-yl)methanone using GC–MS, NMR analysis and a
flash chromatography system. Forensic Science International, 2012.
220(1–3): p. e17-e22.
136. Fowler, F., B. Voyer, M. Marino, J. Finzel, M. Veltri, N.M. Wachter, and
L. Huang, Rapid screening and quantification of synthetic cannabinoids in
herbal products with NMR spectroscopic methods. Analytical Methods,
2015. 7(18): p. 7907-7916.
137. Wick, A., M. Wagner, and T.A. Ternes, Elucidation of the Transformation
Pathway of the Opium Alkaloid Codeine in Biological Wastewater
Treatment. Environmental Science & Technology, 2011. 45(8): p. 3374-
3385.
138. Hagel, J.M. and P.J. Facchini, Plant metabolomics: analytical platforms and
integration with functional genomics. Phytochemistry Reviews, 2008. 7(3):
p. 479-497.
139. Plumb, R.S., C.L. Stumpf, M.V. Gorenstein, J.M. Castro-Perez, G.J. Dear,
M. Anthony, B.C. Sweatman, S.C. Connor, and J.N. Haselden,
Metabonomics: the use of electrospray mass spectrometry coupled to
reversed-phase liquid chromatography shows potential for the screening of
rat urine in drug development. Rapid Communications in Mass
Spectrometry, 2002. 16(20): p. 1991-1996.
140. Zhao, Y.Q., X. Li, X. Liu, P.F. Dong, L.L. Wang, and X.Q. Wang, Research
on water quality analysis model with PCA method and UV absorption
spectra. Guang Pu Xue Yu Guang Pu Fen Xi/Spectroscopy and Spectral
Analysis, 2016. 36(11): p. 3592-3596.
141. Bicchi, C.P., O.M. Panero, G.M. Pellegrino, and A.C. Vanni,
Characterization of Roasted Coffee and Coffee Beverages by Solid Phase
178
Microextraction−Gas Chromatography and Principal Component Analysis.
Journal of Agricultural and Food Chemistry, 1997. 45(12): p. 4680-4686.
142. Cevallos-Cevallos, J.M., J.I. Reyes-De-Corcuera, E. Etxeberria, M.D.
Danyluk, and G.E. Rodrick, Metabolomic analysis in food science: a
review. Trends in Food Science & Technology, 2009. 20(11–12): p. 557-
566.
143. Kallithraka, S., I.S. Arvanitoyannis, P. Kefalas, A. El-Zajouli, E. Soufleros,
and E. Psarra, Instrumental and sensory analysis of Greek wines;
implementation of principal component analysis (PCA) for classification
according to geographical origin. Food Chemistry, 2001. 73(4): p. 501-514.
144. Rodrı́guez-Delgado, M.-Á., G. González-Hernández, J.-E.a. Conde-
González, and J.-P. Pérez-Trujillo, Principal component analysis of the
polyphenol content in young red wines. Food Chemistry, 2002. 78(4): p.
523-532.
145. Snyder, L.R., J.J. Kirkland, and J.W. Dolan, Introduction to modern liquid
chromatography. 2011: John Wiley & Sons.
146. Stachniuk, J., P. Kubalczyk, P. Furmaniak, and R. Głowacki, A versatile
method for analysis of saliva, plasma and urine for total thiols using HPLC
with UV detection. Talanta, 2016. 155: p. 70-77.
147. Li, G., S. Liu, Z. Sun, L. Xia, G. Chen, and J. You, A simple and sensitive
HPLC method based on pre-column fluorescence labelling for multiple
classes of plant growth regulator determination in food samples. Food
Chemistry, 2015. 170: p. 123-130.
148. Ma, J., Z. Yao, L. Hou, W. Lu, Q. Yang, J. Li, and L. Chen, Metal organic
frameworks (MOFs) for magnetic solid-phase extraction of
pyrazole/pyrrole pesticides in environmental water samples followed by
HPLC-DAD determination. Talanta, 2016. 161: p. 686-692.
149. Guiochon, G., N. Marchetti, K. Mriziq, and R.A. Shalliker,
Implementations of two-dimensional liquid chromatography. Journal of
Chromatography A, 2008. 1189(1–2): p. 109-168.
150. Murahashi, T., F. Tsuruga, and S. Sasaki, An automatic method for the
determination of carcinogenic 1-nitropyrene in extracts from automobile
exhaust particulate matter. Analyst, 2003. 128(11): p. 1346-1351.
179
151. Regalado, E.L., J.A. Schariter, and C.J. Welch, Investigation of two-
dimensional high performance liquid chromatography approaches for
reversed phase resolution of warfarin and hydroxywarfarin isomers.
Journal of Chromatography A, 2014. 1363: p. 200-206.
152. Król-Kogus, B., D. Głód, M. Krauze-Baranowska, and I. Matławska,
Application of one- and two-dimensional high-performance liquid
chromatography methodologies for the analysis of C-glycosylflavones from
fenugreek seeds. Journal of Chromatography A, 2014. 1367: p. 48-56.
153. Andrighetto, L.M., P.G. Stevenson, J.R. Pearson, L.C. Henderson, and X.A.
Conlan, DryLab® optimised two-dimensional high performance liquid
chromatography for differentiation of ephedrine and pseudoephedrine
based methamphetamine samples. Forensic Science International, 2014.
244: p. 302-305.
154. Köhne, A.P., U. Dornberger, and T. Welsch, Two-dimensional high
performance liquid chromatography for the separation of complex mixtures
of explosives and their by-products. Chromatographia, 1998. 48(1-2): p. 9-
16.
155. Gilar, M., J. Fridrich, M.R. Schure, and A. Jaworski, Comparison of
Orthogonality Estimation Methods for the Two-Dimensional Separations of
Peptides. Analytical Chemistry, 2012. 84(20): p. 8722-8732.
156. Watson, N.E., J.M. Davis, and R.E. Synovec, Observations on
“Orthogonality” in Comprehensive Two-Dimensional Separations.
Analytical Chemistry, 2007. 79(20): p. 7924-7927.
157. Li, X., D.R. Stoll, and P.W. Carr, Equation for Peak Capacity Estimation in
Two-Dimensional Liquid Chromatography. Analytical Chemistry, 2009.
81(2): p. 845-850.
158. Gilar, M., P. Olivova, A.E. Daly, and J.C. Gebler, Orthogonality of
Separation in Two-Dimensional Liquid Chromatography. Analytical
Chemistry, 2005. 77(19): p. 6426-6434.
159. Lee, M.J., I.-M. Chung, H. Kim, and M.Y. Jung, High resolution LC–ESI-
TOF-mass spectrometry method for fast separation, identification, and
quantification of 12 isoflavones in soybeans and soybean products. Food
Chemistry, 2015. 176: p. 254-262.
180
160. McLafferty, F.W. and F. Tureček, Interpretation of mass spectra. 1993: Mill
Valley, Calif. : University Science Books, [1993] Fourth edition.
161. Saranjit S, T.H., Mallikarjun N, Archana S, Mahendra J, Ravi S, A critical
review on the use of modern sophisticated hyphenated tools in the
characterization of impurities and degradation products. Journal of
Pharmaceutical and Biomedical Analysis, 2012. 69: p. 26.
162. Kohlbacher, O., K. Reinert, C. Gröpl, E. Lange, N. Pfeifer, O. Schulz-
Trieglaff, and M. Sturm, TOPP—the OpenMS proteomics pipeline.
Bioinformatics, 2007. 23(2): p. e191-e197.
163. Sturm, M., A. Bertsch, C. Gropl, A. Hildebrandt, R. Hussong, E. Lange, N.
Pfeifer, O. Schulz-Trieglaff, A. Zerck, K. Reinert, and O. Kohlbacher,
OpenMS - An open-source software framework for mass spectrometry.
BMC Bioinformatics, 2008. 9(1): p. 163.
164. Dioumaeva, I., S.-B. Choi, B. Yong, D. Jones, and R. Arora, Understanding
Orthogonality in Reversed-Phase Liquid Chromatography for Easier
Column Selection and Method Development. Application Note, Agilent
Technologies,
http://www. chem. agilent. com/Library/applications/SI-02425. pdf.
165. Mnatsakanyan, M., P.G. Stevenson, X.A. Conlan, P.S. Francis, T.A.
Goodie, G.P. McDermott, N.W. Barnett, and R.A. Shalliker, The analysis
of café espresso using two-dimensional reversed phase–reversed phase high
performance liquid chromatography with UV-absorbance and
chemiluminescence detection. Talanta, 2010. 82(4): p. 1358-1363.
166. Gritti, F. and G. Guiochon, Adsorption mechanism in reversed-phase liquid
chromatography: Effect of the surface coverage of a monomeric C18-silica
stationary phase. Journal of Chromatography A, 2006. 1115(1–2): p. 142-
163.
167. Gray, M., G.R. Dennis, P. Wormell, R. Andrew Shalliker, and P. Slonecker,
Two dimensional reversed-phase–reversed-phase separations: Isomeric
separations incorporating C18 and carbon clad zirconia stationary phases.
Journal of Chromatography A, 2002. 975(2): p. 285-297.
181
168. Catchpoole, H.J., R. Andrew Shalliker, G.R. Dennis, and G. Guiochon,
Visualising the onset of viscous fingering in chromatography columns.
Journal of Chromatography A, 2006. 1117(2): p. 137-145.
169. Shalliker, R.A., H.J. Catchpoole, G.R. Dennis, and G. Guiochon,
Visualising viscous fingering in chromatography columns: High viscosity
solute plug. Journal of Chromatography A, 2007. 1142(1): p. 48-55.
170. Mayfield, K.J., R.A. Shalliker, H.J. Catchpoole, A.P. Sweeney, V. Wong,
and G. Guiochon, Viscous fingering induced flow instability in
multidimensional liquid chromatography. Journal of Chromatography A,
2005. 1080(2): p. 124-131.
171. Stevenson, P.G., D.N. Bassanese, X.A. Conlan, and N.W. Barnett,
Improving peak shapes with counter gradients in two-dimensional high
performance liquid chromatography. Journal of Chromatography A, 2014.
1337: p. 147-154.
172. Samuelsson, J., R.A. Shalliker, and T. Fornstedt, Viscosity contrast effects
in analytical scale chromatography - Evidence and impact. Microchemical
Journal, 2017. 130: p. 102-107.
173. Ayre, A., D. Varpe, R. Nayak, and N. Vasa, Impurity profiling of
pharmaceuticals. Advance Research in Pharmaceuticals and Biologicals,
2011. 1: p. 76-90.
174. Purcell, D.S.D., Personal correspondance. 2017.
175. Yin, S., X. Zhu, and H.R. Karimi, Quality Evaluation Based on Multivariate
Statistical Methods. Mathematical Problems in Engineering, 2013. 2013: p.
10.
176. Custers, D., B. Krakowska, J.O. De Beer, P. Courselle, M. Daszykowski, S.
Apers, and E. Deconinck, Chromatographic impurity fingerprinting of
genuine and counterfeit Cialis® as a means to compare the discriminating
ability of PDA and MS detection. Talanta, 2016. 146: p. 540-548.
177. Hopfer, H., J. Nelson, T.S. Collins, H. Heymann, and S.E. Ebeler, The
combined impact of vineyard origin and processing winery on the elemental
profile of red wines. Food Chemistry, 2015. 172: p. 486-496.
178. Pap, T.L. and Z. Pápai, Application of a new mathematical function for
describing chromatographic peaks. Journal of Chromatography A, 2001.
930(1–2): p. 53-60.
182
179. Nielsen, N.-P.V., J.M. Carstensen, and J. Smedsgaard, Aligning of single
and multiple wavelength chromatographic profiles for chemometric data
analysis using correlation optimised warping. Journal of Chromatography
A, 1998. 805(1–2): p. 17-35.
180. Stevenson, P.G. and G. Guiochon, Cumulative area of peaks in a
multidimensional high performance liquid chromatogram. Journal of
Chromatography A, 2013. 1308: p. 79-85.
181. Ahuja, S., Assuring quality of drugs by monitoring impurities. Advanced
Drug Delivery Reviews, 2007. 59(1): p. 3-11.
182. Guideline, I.H.T. Impurities in new drug substances Q3A (R2). in
Proceedings of the International Conference on Harmonization of
Technical Requirements for Registration of Pharmaceuticals for Human
Use, Geneva, Switzerland. 2006.
183. Basak, A.K., A.S. Raw, A.H. Al Hakim, S. Furness, N.I. Samaan, D.S. Gill,
H.B. Patel, R.F. Powers, and L. Yu, Pharmaceutical impurities: Regulatory
perspective for Abbreviated New Drug Applications. Advanced Drug
Delivery Reviews, 2007. 59(1): p. 64-72.
184. van Beers, M.M.C. and M. Bardor, Minimizing immunogenicity of
biopharmaceuticals by controlling critical quality attributes of proteins.
Biotechnology Journal, 2012. 7(12): p. 1473-1484.
185. Rice, K.C. and A.H. Newman, Total synthesis of northebaine, normophine,
noroxymorphone enantiomers and derivatives via N-Nor intermediates.
1997, Google Patents.
186. Andrighetto, L.M., N.K. Burns, P.G. Stevenson, J.R. Pearson, L.C.
Henderson, C.J. Bowen, and X.A. Conlan, In-silico optimisation of two-
dimensional high performance liquid chromatography for the determination
of Australian methamphetamine seizure samples. Forensic Science
International, 2016. 266: p. 511-516.
187. Frick, S., J.A. Chitty, R. Kramell, J. Schmidt, R.S. Allen, P.J. Larkin, and
T.M. Kutchan, Transformation of opium poppy (Papaver somniferumL.)
with antisense berberine bridge enzyme gene (anti-bbe) via somatic
embryogenesis results in an altered ratio of alkaloids in latex but not in
roots. Transgenic research, 2004. 13(6): p. 607-613.
183
188. Kelly, S.S., P.M. Glynn, S.J. Madden, and D.H. Grayson, Impurities in a
morphine sulfate drug product identified as 5-(hydroxymethyl)-2-furfural,
10-hydroxymorphine and 10-oxomorphine. Journal of Pharmaceutical
Sciences, 2003. 92(3): p. 485-493.
189. Smith, K.H., T. Bowser, and G.W. Stevens, Performance and Scale-up of
Karr Reciprocating Plate Extraction Columns. Industrial & Engineering
Chemistry Research, 2008. 47(21): p. 8368-8375.
190. Stella, A., K.H. Mensforth, T. Bowser, G.W. Stevens, and H.R.C. Pratt,
Mass Transfer Performance in Karr Reciprocating Plate Extraction
Columns. Industrial & Engineering Chemistry Research, 2008. 47(11): p.
3996-4007.
191. Wiegand, T.J. and F. Lovecchio, Designer drugs: Focus on cathinones (bath
salts) and synthetic cannabinoids (K2 or spice). Emergency Medicine
Reports, 2012. 33(26): p. 305-315.
192. Arnold, C., The new danger of synthetic drugs. The Lancet. 382(9886): p.
15-16.
193. Wilkinson, S.M., S.D. Banister, and M. Kassiou, Bioisosteric Fluorine in
the Clandestine Design of Synthetic Cannabinoids. Australian Journal of
Chemistry, 2015. 68(1): p. 4-8.
194. Merola, G., Z. Aturki, G. D’Orazio, R. Gottardo, T. Macchia, F. Tagliaro,
and S. Fanali, Analysis of synthetic cannabinoids in herbal blends by means
of nano-liquid chromatography. Journal of Pharmaceutical and Biomedical
Analysis, 2012. 71: p. 45-53.
195. Uchiyama, N., M. Kawamura, R. Kikura-Hanajiri, and Y. Goda, URB-754:
A new class of designer drug and 12 synthetic cannabinoids detected in
illegal products. Forensic Science International, 2013. 227(1–3): p. 21-32.
196. Burns, N., Synthetic Cannabinoids: Extraction, identification and
purification, in School of Life and Environmental Sciences. 2013, Deakin
University.
197. Rasanen, I., M. Kyber, I. Szilvay, J. Rintatalo, and I. Ojanperä,
Straightforward single-calibrant quantification of seized designer drugs by
liquid chromatography–chemiluminescence nitrogen detection. Forensic
Science International, 2014. 237: p. 119-125.
184
198. Żwir-Ferenc, A. and M. Biziuk, Solid Phase Extraction Technique - Trends,
Opportunities and Applications. Polish Journal of Environmental Studies,
2006. 15(5): p. 677-690.
199. Grigoryev, A., S. Savchuk, A. Melnik, N.j. Moskaleva, J. Dzhurko, M.
Ershov, A. Nosyrev, A. Vedenin, B. Izotov, I. Zabirova, and V. Rozhanets,
Chromatography–mass spectrometry studies on the metabolism of synthetic
cannabinoids JWH-018 and JWH-073, psychoactive components of
smoking mixtures. Journal of Chromatography B, 2011. 879(15–16): p.
1126-1136.
200. Kavanagh, P., A. Grigoryev, A. Melnik, and A. Simonov, The identification
of the urinary metabolites of 3-(4-methoxybenzoyl)-1-pentylindole (RCS-
4), a novel cannabimimetic, by gas chromatography–mass spectrometry.
Journal of analytical toxicology, 2012. 36(5): p. 303-311.
201. Jang, M., W. Yang, I. Shin, H. Choi, H. Chang, and E. Kim, Determination
of AM-2201 metabolites in urine and comparison with JWH-018 abuse.
International Journal of Legal Medicine, 2014. 128(2): p. 285-294.
202. Rigdon, A., P. Kennedy, and T. Kahler, LC/MS/MS analysis of metabolites
of synthetic cannabinoids JWH-018 and JWH-073 in urine. Clin. Forensic
Toxicol., 2011.
203. Batista, C., M. Berisha, M. Karst, K. Salim, U. Schneider, and R.
Brenneisen, Determination of ajulemic acid and its glucuronide in human
plasma by gas chromatography–mass spectrometry. Journal of
Chromatography B, 2005. 820(1): p. 77-82.
204. de Castro, A., B. Piñeiro, E. Lendoiro, A. Cruz, and M. López-Rivadulla,
Quantification of selected synthetic cannabinoids and Δ9-
tetrahydrocannabinol in oral fluid by liquid chromatography–tandem mass
spectrometry. Journal of Chromatography A, 2013. 1295: p. 99-106.
205. Amaratunga, P., C. Thomas, B.L. Lemberg, and D. Lemberg, Quantitative
measurement of XLR11 and UR-144 in oral fluid by LC–MS-MS. Journal
of analytical toxicology, 2014. 38(6): p. 315-321.
206. Chimalakonda, K.C., C.L. Moran, P.D. Kennedy, G.W. Endres, A. Uzieblo,
P.J. Dobrowolski, E.K. Fifer, J. Lapoint, L.S. Nelson, R.S. Hoffman, L.P.
James, A. Radominska-Pandya, and J.H. Moran, Solid-Phase Extraction
and Quantitative Measurement of Omega and Omega-1 Metabolites of
185
JWH-018 and JWH-073 in Human Urine. Analytical Chemistry, 2011.
83(16): p. 6381-6388.
207. Adcock, J.L., J.M. Terry, C.J. Barrow, N.W. Barnett, D.C. Olson, and P.S.
Francis, Chemiluminescence detectors for liquid chromatography. Drug
Testing and Analysis, 2011. 3(3): p. 139-144.
208. Holland, B.J., P.S. Francis, B. Li, T. Tsuzuki, J.L. Adcock, N.W. Barnett,
and X.A. Conlan, Chemiluminescence detection of cannabinoids and
related compounds with acidic potassium permanganate. Drug Test. Anal.,
2012. 4(7-8): p. 675-679.
209. McDermott, G.P., P. Jones, N.W. Barnett, D.N. Donaldson, and P.S.
Francis, Stable Tris(2,2′-bipyridine)ruthenium(III) for Chemiluminescence
Detection. Analytical Chemistry, 2011. 83(13): p. 5453-5457.
210. Ong, E.S., Extraction methods and chemical standardization of botanicals
and herbal preparations. Journal of Chromatography B, 2004. 812(1–2): p.
23-33.
211. Oulton, S. Scientific working group for the analysis of seized drugs. 2017
[cited 2016 16/04/2016]; Available from: http://www.swgdrug.org/.
212. Pandohee, J., B.J. Holland, B. Li, T. Tsuzuki, P.G. Stevenson, N.W.
Barnett, J.R. Pearson, O.A.H. Jones, and X.A. Conlan, Screening of
cannabinoids in industrial-grade hemp using two-dimensional liquid
chromatography coupled with acidic potassium permanganate
chemiluminescence detection. Journal of Separation Science, 2015. 38(12):
p. 2024-2032.
213. Dash, S., S. Patel, and B.K. Mishra, Oxidation by permanganate: synthetic
and mechanistic aspects. Tetrahedron, 2009. 65(4): p. 707-739.
214. Adcock, J.L., P.S. Francis, T.A. Smith, and N.W. Barnett, The characteristic
red chemiluminescence from reactions with acidic potassium
permanganate: further spectroscopic evidence for a manganese(ii) emitter.
Analyst, 2008. 133(1): p. 49-51.
215. Chen, G.N., F.X. Huang, X.P. Wu, Z.F. Zhao, and J.P. Duan,
Chemiluminescence determination of melatonin and some of its derivatives
using potassium permanganate and formaldehyde system. Analytical and
Bioanalytical Chemistry, 2003. 376(6): p. 873-878.
186
216. Han, S., Chemiluminescence determination of indole derivatives in human
body fluids and soil by flow injection analysis using potassium
permanganate. Microchimica Acta, 2010. 168(1): p. 169-175.
217. Nie, F., J. Lu, Y. He, and J. Du, Determination of indomethacin in urine
using molecule imprinting-chemiluminescence method. Talanta, 2005.
66(3): p. 728-733.
218. Hindson, C.M., P.S. Francis, G.R. Hanson, J.L. Adcock, and N.W. Barnett,
Mechanism of Permanganate Chemiluminescence. Analytical Chemistry,
2010. 82(10): p. 4174-4180.
219. Slezak, T., Z.M. Smith, J.L. Adcock, C.M. Hindson, N.W. Barnett, P.N.
Nesterenko, and P.S. Francis, Kinetics and selectivity of permanganate
chemiluminescence: A study of hydroxyl and amino disubstituted benzene
positional isomers. Analytica Chimica Acta, 2011. 707(1–2): p. 121-127.
220. Smith, Z.M., J.M. Terry, N.W. Barnett, L.J. Gray, D.J. Wright, and P.S.
Francis, Enhancing permanganate chemiluminescence detection for the
determination of glutathione and glutathione disulfide in biological
matrices. Analyst, 2014. 139(10): p. 2416-2422.
221. Smith, Z.M., J.M. Terry, C.M. Hindson, N.W. Barnett, and P.S. Francis,
On-line generation of a colloidal manganese(IV) reagent for
chemiluminescence detection. Microchemical Journal, 2013. 111: p. 67-73.
222. Dronova, M., E. Smolianitski, and O. Lev, Electrooxidation of New
Synthetic Cannabinoids: Voltammetric Determination of Drugs in Seized
Street Samples and Artificial Saliva. Analytical Chemistry, 2016. 88(8): p.
4487-4494.
223. Knight, A.W. and G.M. Greenway, Relationship between structural
attributes and observed electrogenerated chemiluminescence (ECL) activity
of tertiary amines as potential analytes for the tris(2,2-
bipyridine)ruthenium(II) ECL reaction. A review. Analyst, 1996. 121(11):
p. 101R-106R.
224. Uchikura, K. and M. Kirisawa, Electrochemiluminescence Determination
of D-, L-Tryptophan Using Ligand-Exchange High-Performance Liquid
Chromatography. Analytical Sciences, 1991. 7(6): p. 971-973.
225. Mervartová, K., M. Polášek, and J.M. Calatayud, Sequential injection
analysis (SIA)-chemiluminescence determination of indomethacin using
187
tris[(2,2′-bipyridyl)]ruthenium(III) as reagent and its application to
semisolid pharmaceutical dosage forms. Analytica Chimica Acta, 2007.
600(1–2): p. 114-121.
226. Tortajada-Genaro, L.A. and P. Campíns-Falcó, Multivariate standardisation
for non-linear calibration range in the chemiluminescence determination of
chromium. Talanta, 2007. 72(3): p. 1004-1012.
227. Kelly, T.J., C.W. Spicer, and G.F. Ward, An assessment of the luminol
chemiluminescence technique for measurement of NO2 in ambient air.
Atmospheric Environment. Part A. General Topics, 1990. 24(9): p. 2397-
2403.
228. Logan, B.K., L.E. Reinhold, A. Xu, and F.X. Diamond, Identification of
Synthetic Cannabinoids in Herbal Incense Blends in the United States.
Journal of Forensic Sciences, 2012. 57(5): p. 1168-1180.
229. Kneisel, S., F. Westphal, P. Bisel, V. Brecht, S. Broecker, and V. Auwärter,
Identification and structural characterization of the synthetic cannabinoid
3-(1-adamantoyl)-1-pentylindole as an additive in ‘herbal incense’. Journal
of Mass Spectrometry, 2012. 47(2): p. 195-200.
230. Zuba, D., B. Byrska, and M. Maciow, Comparison of “herbal highs”
composition. Analytical and Bioanalytical Chemistry, 2011. 400(1): p. 119-
126.
231. Marino, M.A., B. Voyer, R.B. Cody, A.J. Dane, M. Veltri, and L. Huang,
Rapid Identification of Synthetic Cannabinoids in Herbal Incenses with
DART-MS and NMR. Journal of Forensic Sciences, 2016. 61: p. S82-S91.
232. Ernst, L., H.-M. Schiebel, C. Theuring, R. Lindigkeit, and T. Beuerle,
Identification and characterization of JWH-122 used as new ingredient in
“Spice-like” herbal incenses. Forensic Science International, 2011. 208(1–
3): p. e31-e35.
233. Namespetra, A.M., D.A. Hirsh, M.P. Hildebrand, A.R. Sandre, H. Hamaed,
J.M. Rawson, and R.W. Schurko, 35Cl solid-state NMR spectroscopy of
HCl pharmaceuticals and their polymorphs in bulk and dosage forms.
CrystEngComm, 2016. 18(33): p. 6213-6232.
234. Hofstetter, A. and L. Emsley, Positional Variance in NMR Crystallography.
Journal of the American Chemical Society, 2017. 139(7): p. 2573-2576.
188
235. Bardet, M. and A. Pournou, Chapter Two - NMR Studies of Fossilized
Wood, in Annual Reports on NMR Spectroscopy, A.W. Graham, Editor.
2017, Academic Press. p. 41-83.
236. Ashbrook, S.E., D.M. Dawson, and V.R. Seymour, Recent developments in
solid-state NMR spectroscopy of crystalline microporous materials.
Physical Chemistry Chemical Physics, 2014. 16(18): p. 8223-8242.
237. Lee, D., G. Monin, N.T. Duong, I.Z. Lopez, M. Bardet, V. Mareau, L.
Gonon, and G. De Paëpe, Untangling the Condensation Network of
Organosiloxanes on Nanoparticles using 2D 29Si–29Si Solid-State NMR
Enhanced by Dynamic Nuclear Polarization. Journal of the American
Chemical Society, 2014. 136(39): p. 13781-13788.
238. Baldus, M., Solid-State NMR Spectroscopy: Molecular Structure and
Organization at the Atomic Level. Angewandte Chemie International
Edition, 2006. 45(8): p. 1186-1188.
239. Berendt, R.T., D.M. Sperger, E.J. Munson, and P.K. Isbester, Solid-state
NMR spectroscopy in pharmaceutical research and analysis. TrAC Trends
in Analytical Chemistry, 2006. 25(10): p. 977-984.
240. Paradowska, K. and I. Wawer, Solid-state NMR in the analysis of drugs and
naturally occurring materials. Journal of Pharmaceutical and Biomedical
Analysis, 2014. 93: p. 27-42.
241. Schaefer, J. and E.O. Stejskal, Carbon-13 nuclear magnetic resonance of
polymers spinning at the magic angle. Journal of the American Chemical
Society, 1976. 98(4): p. 1031-1032.
242. Albert, K. and E. Bayer, Characterization of bonded phases by solid-state
NMR spectroscopy. Journal of Chromatography A, 1991. 544: p. 345-370.
243. Rossini, A.J., A. Zagdoun, M. Lelli, A. Lesage, C. Coperet, and L. Emsley,
Dynamic nuclear polarization surface enhanced NMR spectroscopy.
Accounts of chemical research, 2013. 46(9): p. 1942-1951.
244. Bower, P.V., E.A. Louie, J.R. Long, P.S. Stayton, and G.P. Drobny, Solid-
State NMR Structural Studies of Peptides Immobilized on Gold
Nanoparticles. Langmuir, 2005. 21(7): p. 3002-3007.
245. Linton, R.W., M.L. Miller, G.E. Maciel, and B.L. Hawkins, Surface
characterization of chemically modified (trimethylsily) silicas by 29Si solid
189
state NMR, XPS, and IR photoacoustic spectroscopy. Surface and Interface
Analysis, 1985. 7(4): p. 196-203.
246. Pallister, P.J., S.C. Buttera, and S.T. Barry, Quantitative Surface Coverage
Calculations via Solid-State NMR for Thin Film Depositions: A Case Study
for Silica and a Gallium Amidinate. The Journal of Physical Chemistry C,
2014. 118(3): p. 1618-1627.
247. Ladizhansky, V. and S. Vega, Doping of CdS Nanoparticles by Co2+ Ions
Studied by NMR. The Journal of Physical Chemistry B, 2000. 104(22): p.
5237-5241.
248. Jäger, C., T. Welzel, W. Meyer-Zaika, and M. Epple, A solid-state NMR
investigation of the structure of nanocrystalline hydroxyapatite. Magnetic
Resonance in Chemistry, 2006. 44(6): p. 573-580.
249. Eckman, R.R., P.M. Henrichs, and A.J. Peacock, Study of Polyethylene by
Solid State NMR Relaxation and Spin Diffusion. Macromolecules, 1997.
30(8): p. 2474-2481.