Content uploaded by Arno Hazekamp
Author content
All content in this area was uploaded by Arno Hazekamp on Oct 03, 2016
Content may be subject to copyright.
ORIGINAL RESEARCH Open Access
Cannabis:
From Cultivar to Chemovar II—A Metabolomics
Approach to Cannabis Classification
Arno Hazekamp,
1,
*Katerina Tejkalova
´,
2
and Stelios Papadimitriou
2
Abstract
Introduction: There is a large disparity between the ‘‘cultural’’ language used by patients using cannabis for self-
medication and the ‘‘chemical’’ language applied by scientists to get a deeper understanding of cannabis effects in
laboratory and clinical studies. The distinction between Sativa and Indica types of cannabis, and the different bio-
logical effects associated with them, is a major example of this. Despite the widespread use of cannabis by self-
medicating patients, scientific studies are yet to identify the biochemical markers that can sufficiently explain dif-
ferences between cannabis varieties.
Methods: A metabolomics approach, combining detailed chemical composition data with cultural information
available for a wide range of cannabis samples, can help to bridge the existing gap between scientists and patients.
Such an approach could be helpful for decision-making, for example, when identifying which varieties of cannabis
should be made legally available under national medicinal cannabis programs. In our study, we analyzed 460 can-
nabis accessions obtained from multiple sources in The Netherlands, including hemp- and drug-type cannabis.
Results: Based on gas chromatography analysis of 44 major terpenes and cannabinoids present in these sam-
ples, followed by Multivariate Data Analysis of the resulting chromatographic data, we were able to identify the
cannabis constituents that may act as markers for distinction between Indica and Sativa. This information was
subsequently used to map the current chemical diversity of cannabis products available within the Dutch me-
dicinal cannabis program, and to introduce a new variety missing from the existing product range.
Conclusion: This study represents the analysis of the widest range of cannabis constituents published to date.
Our results indicate the usefulness of a metabolomics approach for chemotaxonomic mapping of cannabis va-
rieties for medical use.
Keywords: cannabinoids; cannabis; gas chromatography; multivariate data analysis; terpenes; varieties
Introduction
Research into herbal cannabis poses serious challenges to
modern medicine, which operates mainly according to
the ‘‘single compound–single target’’ paradigm of phar-
macology. Although it was once believed that THC (see
Table1forabbreviations)wastheuniqueactiveprinci-
ple found in cannabis, it is becoming clear that a much
wider range of cannabis constituents may be involved
in its various therapeutic effects. Currently, many dif-
ferent subtypes of cannabis are known to exist, and
the high number of (potential) active components
significantly complicates a conventional reductionist
approach using analytical chemistry, animal studies,
and clinical trials, where typically a single active in-
gredient is identified before development of a final
medicine is possible. With the recent growth in me-
dicinal use of cannabis, the need to clearly distin-
guish between various cannabis plants and their
1
Department of Research and Education, Bedrocan BV, Veendam, The Netherlands.
2
Natural Products Laboratory, Institute of Biology, Leiden University, The Netherlands.
*Address correspondence to: Arno Hazekamp, PhD, Department of Research and Education, Bedrocan BV, Veendam 9640CA, The Netherlands, E-mail: a.hazekamp@bedrocan.com
ªArno Hazekamp et al. 2016; Published by Mary Ann Liebert, Inc. This Open Access article is distributed under the terms of the Creative Commons
License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly credited.
Cannabis and Cannabinoid Research
Volume 1.1, 2016
DOI: 10.1089/can.2016.0017
Cannabis and
Cannabinoid Research
202
expected (therapeutic) effects has become more im-
portant than ever.
Medicinal cannabis may have therapeutic effects on
various illnesses ranging from chronic pain and multiple
sclerosis (MS), to epilepsy and anxiety.
1,2
An obvious
question is how the chemical composition of cannabis re-
flects these various medicinal effects, and what types of
cannabis should consequently be made available to pro-
vide patients with a full spectrum of therapeutic benefits.
Already more than 700 different cultivated varieties (cul-
tivars) of cannabis have been cataloged
3
and many more
are thought to exist, each one with a potentially different
effect on body and mind. Although pharmacopoeia
monographs for the chemical analysis of cannabis have
been developed in several countries,
4–6
these are intended
for potency, quality, and adulteration issues only. They
are not designed to map out the complex chemical vari-
ations found among different cannabis products. A bet-
ter classification of cannabis varieties, and the chemical
differences between them, would certainly promote fur-
ther implementation of cannabis-based products into
(pre)clinical research and modern medicine.
Science-based classification systems
The scientific debate about classification of the cannabis
species has been going onfor centuries. According to cur-
rent scientific consensus, Cannabis is monotypic and
consists only of a single species Cannabis sativa L.,as
originally described by Leonard Fuchs in the 16th cen-
tury.
3
Within this species, two important subdivisions
are commonly made. One of them recognizes drug-
type versus fiber-type cannabis based on the intended
use of the plant, and is mostly relevant for legal purposes.
The other subdivision is based on botanical principles
and identifies Sativa versus Indica types of cannabis,
both regarded as a subspecies of Cannabis sativa L.
7
According to the botanical description of cannabis,
Sativa types of cannabis were originally grown in the
Western world on an industrial scale for fiber, oil, and
animal feedstuff. They are characterized by tall growth
with few, widely spaced, branches and long, thin leaves.
In contrast, plants of the Indica type originated in South
Asia and were known historically as Indian hemp. They
are characterized by shorter bushy plants and broader
leaves, typically maturing relatively fast. The two groups
tend to have a different smell, which may reflect a differ-
ent profile of fragrant terpenes. Most cannabis plants
that are currently commercially available are in fact a
hybrid (cross-breed) of Sativa and Indica ancestors.
Cannabis-type ruderalis is sometimes also recognized
as a separate subspecies. It is a smaller and ‘‘weedy’’
plant originally from Central Russia.
7
Various scientific attempts have been made to classify
cannabis plants based on their cannabinoid composition,
as outlined in more detail in our previous article.
8
For fo-
rensic and legal purposes, the most important classifica-
tion is that of the drug type (marijuana) versus the
fiber type (hemp), with an emphasis on the total THC
Table 1. List of Analyzed Terpenes and Cannabinoids
RT (min) RRT Compound
Internal standard 10.64 0.427 1-Octanol
Monoterpenes 6.37 0.256 Alpha-2-pinene
7.59 0.304 Beta-2-pinene
7.94 0.318 Myrcene
8.38 0.336 Alpha-phellandrene
8.60 0.345 Delta-3-carene
9.18 0.368 Beta-phellandrene
(1)
9.23 0.370 R-limonene
(1)
9.33 0.374 Cineol
9.88 0.396 Cis-ocimene
10.29 0.413 Gamma-terpinene
11.39 0.457 Terpinolene
11.80 0.473 ()-Linalool
12.36 0.496 Beta-fenchol
12.67 0.508 Cis-sabinene hydrate
13.70 0.549 Camphor
14.46 0.580 Borneol
15.49 0.621 Alpha-terpineol
Sesquiterpenes 23.86 0.957 Cis-bergamotene
24.93 1.000 Beta-caryophyllene
25.57 1.026 Trans-bergamotene
25.68 1.030 Alpha-guaiene
25.70 1.031 Aromadendrene
26.28 1.054 Alpha-humulene
26.41 1.059 Trans-beta-farnesene
28.44 1.141 Gamma-selinene
28.56 1.146 Delta-guaiene
29.45 1.181 Gamma-cadinene
29.70 1.191 Eudesma-3,7(11)-diene
30.27 1.214 Gamma-elemene
30.69 1.231 Nerolidol
(2)
31.23 1.253 Trans-beta-caryophyllene
31.29 1.255 Beta-caryophyllene oxide
31.81 1.276 Guaiol
32.62 1.308 Gamma-eudesmol
33.67 1.351 Beta-eudesmol
34.89 1.400 Alpha-bisabolol
Cannabinoids 54.46 2.185 Tetrahydrocannabivarin (THCV)
57.29 2.298 Cannabidiol (CBD)
57.58 2.310 Cannabichromene (CBC)
58.83 2.360 Cannabigerol-monomethylether
(CBGM)
59.18 2.374 Delta-8-tetrahydrocannabinol
(delta 8-THC)
(3)
59.89 2.402 Delta-9-tetrahydrocannabinol
(THC)
(3)
61.37 2.461 Cannabigerol (CBG)
61.57 2.469 Cannabinol (CBN)
(3)
(1): beta-phellandrene and R-limonene were combined for analysis be-
cause of chromatographic peak overlap. (2): nerolidol may in fact be
alpha-gurjunene. (3): delta-8-THC, CBN, and THC were combined for
final analysis to one value for ‘‘THC total.’’
GC, gas chromatography; RT, GC retention time in minutes; RRT, rela-
tive retention time compared to beta-caryophyllene.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
203
content in the flowers of the plant. Until recently, canna-
bis products used for medicinal purposes in official pro-
grams all belonged to the drug type, because of their high
content of the biologically active THC. However, it is
becoming increasingly clear that multiple constituents
may be involved in the overall effect of the drug.
9
This includes the cannabinoids CBD,
10,11
CBG,
12,13
and THCV,
14,15
as well as a variety of terpenes.
16
Terpenes are volatile compounds responsible for the
typical smell and taste of cannabis. More than 120 dif-
ferent types have been identified in cannabis and their
relative composition can vary widely between varie-
ties.
17,18
Terpenes present in cannabis have a wide
range of known biological effects, and some may be in-
volved in regulating and/or modifying the effects of
THC and other cannabinoids.
16
Expanding our scien-
tific understanding of the therapeutic benefits of ter-
penes is a budding new frontier of medical cannabis
research. Nevertheless, no currently used cannabis clas-
sification system takes the terpenes into account.
Since the publication of the draft genome of canna-
bis,
19
various studies have looked into the genetic anal-
ysis of cannabis as a means of identifying distinct
(sub)groups.
20,21
Such studies have been able to clearly
differentiate hemp from marijuana-type cannabis,
21
but they find only a moderate correlation between
the ancestry of cannabis strains, as reported by breed-
ers, and the ancestry inferred from their DNA. As a re-
sult, the genetic identity of cannabis products could not
be reliably inferred by its vernacular name or by its
reported ancestry as supplied by cannabis growers.
Vernacular classification
Over the years, illicit growers of marijuana have created
an enormous range of cultivated varieties, also popu-
larly known as cultivars or strains. These are com-
monly distinguished, by plant breeders, recreational
users, and medical cannabis patients alike, through
the use of popularized vernacular names such as
White Widow,Northern Lights,AK-47,orAmnesia
Haze. The most common way to categorize such strains
is based on plant morphology (phenotype), which con-
siders features such as leaf shape, plant height, color,
smell, and speed of growth. This has resulted in a ver-
nacular system of distinguishing Sativa from Indica
types of cannabis, which developed independent from
scientific and taxonomic classification systems.
3
According to the vernacular system, Sativa strains are
usually characterized as uplifting and energetic. The ef-
fects are mostly cerebral (‘‘head-high’’), also described
as spacey or hallucinogenic. This type gives a feeling
of optimism and well-being. In contrast, Indica strains
are primarily described as calming and grounding
(‘‘body-high’’). This type is said to cause relaxation,
stress relief, and an overall sense of serenity.
22
Because of limited guidance and support from the
medical community, self-medicating users of cannabis
often adopt the vocabulary of cannabis subculture to
navigate their search for cannabis medicine.
23
When
buying cannabis for recreational as well as for medici-
nal reasons, consumers use the vernacular Sativa and
Indica labels as a means to identify their preferred
products. Unfortunately, it is largely unclear how
these labels reflect an actual difference in chemical
composition, and how they are related to the scientifi-
cally accepted botanical classification of cannabis culti-
vars, as described above. A major issue with vernacular
classification is that there are no generally agreed stan-
dards or enforced regulations for variety labeling and
naming. As a result, the Sativa/Indica distinction made
by botanists and taxonomists may be very different
from the one applied by consumers of cannabis.
22
Several studies have challenged the usefulness of the
current Sativa/Indica classification system.
3
For exam-
ple, in a recent study, almost 500 cannabis samples
available to California patients were analyzed for chem-
ical composition.
24
The study concluded that popular
product names as defined by the authoritative Leafly
database (www.leafly.com) were no guarantee to ob-
tain a product with a reproducible chemical composi-
tion, and that the provided product name was not a
reliable indicator of THC potency or chemical profile.
In fact, samples collected under the same name often
did not even look similar in appearance. Alternative
classifications have been proposed, such as Broad
Leaflet versus Narrow Leaflet drug types.
20,25
Another
recent article suggested a complete overhaul of the
current botanical classification based on reinterpreta-
tion of historical taxonomic records.
26
From cultivar to chemovar
For medicinal users of cannabis, the main challenge is
to find a product with a chemical profile that matches
their pharmacological need for treatment. The name of
the cannabis product serves merely as a means to iden-
tify and purchase the desired product. What is needed
is a practical manner to more directly visualize the
chemical diversity present within the many cannabis
products offered, and to make sure that the full range
of diversity is accessible through legally available,
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
204
high-quality products under national programs safe-
guarding quality and consistency.
A comprehensive overview of chemical diversity can
help medicinal users and their physicians to successfully
transition from a beneficial cannabis product obtained
through illicit markets, to a comparable but high-quality
variety that is available through official national pro-
grams. It may also help these national programs to nar-
row down the search for beneficial cannabis cultivars to
be legally produced and introduced into healthcare.
Such an approach fits exactly within the research field
of metabolomics, which is recognized as a scientific
methodology to better understand the complex effects
of medicinal plants containing multiple active ingredi-
ents. A metabolomics approach would simultaneously
identify and quantify all major cannabinoids and ter-
penes present in various cannabis types, and then use
powerful statistical tools such as Multivariate Data Anal-
ysis to create a ‘‘map’’ of chemical diversity.
This methodology has already been successfully ap-
plied to cannabis for differentiation of cannabis products
on a small scale as well as for quality control.
8,27
With
this approach, it may be possible to move away from
the current system of cannabis cultivars,withoften
vague and unsubstantiated characteristics dominated
by Sativa/Indica labeling, toward a new classification
based on chemical varieties (or ‘‘chemovars’’) with a com-
plex, but well-defined and reproducible chemical profile.
Goal of our study
In this article, we address the question whether cannabis
vernacular product names can be replaced with chemical
profilesasameansofclassifyingcannabisproductsinto
distinct groups. A metabolomics approach was used to in-
vestigate the chemical variation present within 460 acces-
sions of marijuana as well as hemp collected in The
Netherlands. We were particularly interested in the differ-
ences in composition between the Indica and Sativa types.
Samples were chemically profiled for 44 different major
cannabinoids and terpenes using gas chromatography
with flame ionization detection (GC/FID), followed by sta-
tistical evaluation of the chromatographic data. The study
results indicate the usefulness of the chemovar approach
(full chemical analysis for chemotaxonomic mapping of
cannabis varieties) and may assist in a better identification
of cannabis products with a potential for medicinal use.
To highlight the feasibility of our proposed chemovar
mapping, we describe an example from the Dutch me-
dicinal cannabis program where the experiences of
cannabis-using patients were combined with our chem-
ical profiling approach to introduce a new standardized
cannabis variety for Dutch patients.
Materials and Methods
Solvents and chemicals
Authentic standards were available for the following ter-
penes: alpha-pinene, beta-pinene, beta-caryophyllene,
ocimene, alpha-phellandrene, R-(+)-limonene, gamma-
terpinene, (+)-borneol, () caryophyllene oxide, ()-
trans-caryophyllene, cineole, fenchol, alpha-bisabolol,
camphor, myrcene, farnesene (mixture of isomers), S-()-
limonene, gamma-terpineol, phytol, carvacrol, and (+)-
aromadendrene were obtained from Sigma-Aldrich;
alpha-humulene, terpineol, ()-linalool, and trans-
nerolidol were from Fluka; beta-eudesmol and geraniol
were from Chromadex; delta-3-carene was from Roth.
Calibrated standards were available for the following
cannabinoids: delta-9-tetrahydrocannabinol (THC),
cannabidiol (CBD), cannabigerol (CBG), cannabichro-
mene (CBC), trans-()-delta-9-tetrahydrocannabivarin
(THCV), and cannabinol (CBN). They were purified
and quantified as previously described.
8,28,29
A stan-
dard for cannabigerol-monomethylether (CBGM) was
not available.
All organic solvents and 1-octanol (used as internal
standard [IS]) were of analytical reagent grade or
HPLC grade (Sigma-Aldrich).
Sample collection
A total of 460 accessions were collected from three dif-
ferent sources (coffeeshops, Bedrocan, HempFlax) as
described below. Vernacular names and presumed
botanical type (Sativa, Indica, hybrid, hemp) were
recorded as provided by the source. All samples were
kept in a freezer at 20C until analyzed.
Cannabis cultivation and sales are formally illegal
under Dutch law. However, through the famous outlets
known as ‘‘coffeeshops’’ the sales of small quantities of
cannabis are tolerated (condoned) under strict condi-
tions. There are currently about 600 coffeeshops in
The Netherlands, with the majority located in the big-
ger cities.
30
We collected a total of 56 accessions from
various coffeeshops present in several Dutch munici-
palities. One gram of each product was purchased
and stored in the package provided (typically a small
plastic zip-lock bag). In general, we requested to pur-
chase the most typical Sativa-dominant and Indica-
dominant products available in the shop. No verifiable
phenotypic information about the original plants is
available for these samples.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
205
Bedrocan BV is the official cultivator of pharmaceutical-
grade cannabis within the Dutch medicinal cannabis pro-
gram. A total of 37 different accessions were obtained from
its breeding program. Included were six standardized vari-
eties currently available to patients on prescription (official
product names: Bedrocan, Bedrobinol, Bediol, Bedica,
Bedrolite, and Bedropuur). Voucher specimens have
been deposited at the Naturalis National Herbarium in
Leiden, The Netherlands (L) under sample codes ‘‘Haze-
kamp 1–6.’’ All accessions were grown from female clones
under strictly standardized indoor conditions. Flowertops
were harvested and air-dried for 1 week under controlled
temperature and humidity. Final samples were delivered
in airtight, triple-laminate aluminum foil bags. Detailed
descriptions and pictures of the plants are available.
HempFlax BV produces semi-finished and finished
products made from ecologically grown fiber hemp
and flax. Under their current research license, the com-
pany is propagating a wide range of cannabis varieties,
including hemp, Sativa, and Indica types. Seeds of these
accessions are sold through the Dutch company Sensi
Seed Bank. From this source, we received a total of
367 accessions in the form of dried flowers. Plants
were grown under indoor conditions. Samples were de-
livered in plastic zip-lock bags. Detailed descriptions
and pictures of these plants are available.
Sample extraction
Samples were homogenized by manually cutting the
plant material in smaller pieces with scissors. Of each
sample, 500 mg (accuracy –5 mg) was weighed and
transferred to a 50-mL plastic serum tube. To each
tube, 40 mL of ethanol (absolute) was added, and the
tube was agitated for 10 min with a mechanical shaker
(Yellow Line Orbital Shaker OS 2 Basic; IKA GmbH) at
300 rpm. The tube was then centrifuged and clear super-
natant was poured into a 100-mL volumetric flask. For
exhaustive extraction, the procedure was repeated twice
more with 25 mL of ethanol, and supernatants were com-
bined. At this point, 200 lLofIS,consistingofa1%so-
lution of 1-octanol in ethanol, was added to each extract,
resulting in a final IS concentration of 0.002% (v/v).
Finally, volume was adjusted to 100 mL with ethanol
and mixed well. The solution was centrifuged to remove
solid particles and stored at20C until analyzed by GC.
All accessions were analyzed in duplicate.
GC analysis
An Agilent GC 6890 series (Agilent Technologies, Inc.)
equipped with a 7683 autosampler and an FID was
used for simultaneous analysis of monoterpenes, sesqui-
terpenes, and cannabinoids, as previously described.
8
The instrument was equipped with a DB5 column
(30 m length, 0.25mm internal diameter, film thickness
0.25 lm; J&W Scientific, Inc.). The injector temperature
was 230C, with an injection volume of 1 lL, a split ratio
of 1:20, and a carrier gas (N
2
) flow rate of 1.2 mL/min.
The temperature gradient started at 60C and increased
at a rate of 3C/min until 240C, which was held for
5 min making a total run time of 65 min/sample. The
FID detector temperature was set to 250C. The GC-
FID was controlled by Agilent GC Chemstation soft-
ware version (rev. B.04.01).
In each sample chromatogram, typically 30–40 chro-
matographic peaks were visible. All peaks were manually
integrated and resulting peak area values were entered
into an Excel datasheet. Raw integration data were
then corrected for IS (1-octanol) peak area and used
for principal component analysis (PCA) as described
below. It should be noted that for each sample, the
peak areas for CBN, delta-8-THC, and THC, were com-
bined to one single value for ‘‘THC total.’’ Because CBN
and delta-8-THC are known degradation products of
THC (resulting from long-term storage and aging) and
the cannabis plant does not originally produce these
degradation products as part of its metabolism, they
are considered to be part of the actual (original) content
of THC. In general, peak areas of CBN and delta-8-THC
were present in trace amounts only.
Peak identification and quantification
To confirm peak identification, selected samples were
further analyzed by GC-MS using a single quadrupole
MS-detector in total ion count mode as previously de-
scribed.
27
Compounds were compared based on their
mass spectra and retention times with authentic stan-
dards as well as literature reports.
31–36
In addition,
the NIST library was used to assist in compound iden-
tification (version 2.0f; Standard Reference Data Pro-
gram of the National Institute of Standards and
Technology, as distributed by Agilent Technologies).
For quantitative analysis, peak area values were
quantified (in mg/g of plant material) with the use
of calibration curves. Monoterpenes, sesquiterpenes,
and cannabinoids present in the samples were quanti-
fied by using calibrated standards of beta-pinene,
alpha-humulene, and CBD, respectively, comparable
to our method previously described.
8
Each calibra-
tion curve consisted of four different concentration
levels in the range of 0.005–0.1 mg/mL (beta-pinene,
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
206
alpha-humulene) or 0.25–1.0 mg/mL (CBD) dissolved
in ethanol. Calibration curves were regularly prepared
throughout the duration of the study. The resulting
quantitative data were not corrected for residual
moisture content of the samples.
Multivariate data analysis
After correcting for IS and averaging of duplicate ana-
lyses, peak integrals from the GC-FID chromatograms
were used for Multivariate Data Analysis with the aid
of SIMCAP+version 13.0.3 software (Umetrics AB).
All values were scaled with unit variance before model-
ing, which gives equal weight to all variables.
First, unsupervised analysis was performed by applying
PCA to identify clusters of samples that have similar
chemical composition. PCA is a statistical way of identi-
fying patterns in complex data sets, by turning the data set
into linearly uncorrelated values, which are the principal
components (PCs).
37
The resulting plot highlights the
major similarities and differences between the analyzed
samples.
38
With this method, there is no a-priori expecta-
tion of the nature of such clusters, and we leave it up to
the computer program to tell usifclustersofdatapoints
actually exist. For this analysis, we used GC data of all 460
accessions. The first two principal components (PC1 and
PC2) were visualized in a scatter plot, in combination
with its accompanying loading plot, to show the contribu-
tion of each chemical component to the diversity ob-
served in the sample collection.
Next, a supervised analysis was performed by applying
orthogonal projection to latent structure-discriminant
analysis (OPLS-DA) on the GC data obtained from the
131 accessions in our data set that were labeled as pure
Indica or Sativa. OPLS-DA is a form of partial least-
squares regression analysis (a dimensionality reduction
technique) combined with discriminant analysis (which
is a form of classification). OPLS-DA is a further develop-
ment from PCA and a method of choice when working
with a data set consisting of known distinct classes (in
our case, Sativa versus Indica).
39
This method shows
which variables are responsible for class discrimination
in an easy to interpret 2D visualization.
40
In addition, for a better understanding of the quantita-
tive differences between Sativa, Indica, and hemp types of
cannabis, GC data were quantified with the use of calibra-
tion curves to determine the absolute content (in milli-
grams) of each component per gram of plant material.
Statistical analysis was done by T-test to identify signifi-
cant differences in cannabinoid or terpene composition
between these three cannabis types. Finally, PCA map-
ping of the entire cannabinoid and terpene profile was
shown to be a useful way to identify missing cannabis
types within the Dutch medicinal cannabis program.
Results and Discussion
Cannabis sampling and analysis
In the present study, we chemically analyzed 460 differ-
ent cannabis accessions from various sources, represent-
ing one of the largest cannabis sample sets analyzed to
date. Based on information obtained from suppliers, ac-
cessions were labeled as Hemp (n=121), Sativa (n=68),
Indica (n=63), or Hybrid (n=208). With the use of
GC analysis, 44 different compounds could be posi-
tively identified with a single chromatographic run,
including 17 monoterpenes, 19 sesquiterpenes, and 8
cannabinoids (Table 1). It should be noted that during
GC analysis, all acidic cannabinoids (e.g., THC-acid
and CBD-acid) are fully converted to their neutral
counterparts (THC, CBD, etc.) as a result of the high
temperatures (up to 250C) applied during injection
and separation.
Since our previous study,
8
we were able to extend our
list of positively identified terpenes from 20 to 36, by
more closely studying the compounds previously identi-
fied by Ross and ElSohly
33
and Hillig,
34
as well as by the
use of newly acquired standards and performing more
detailed GC-MS analysis of unresolved peaks. Only
one of the newly added compounds could not be un-
equivocally identified: the compound we putatively
identified as nerolidol (in accordance with Ross and
ElSohly
33
) may perhaps be alpha-gurjunene.
34
Unfortu-
nately, we did not have the pure standard for alpha-
gurjunene needed to distinguish between these two.
Because of lack of resolution between the chromato-
graphic peaks of beta-phellandrene and R-limonene,
these two compounds were analyzed as a single peak.
Compared to our previous study on cannabis compo-
sition,
8
we made one modification to the extraction and
analysis protocol, that is, the addition of an IS in the
form of 1-octanol. In each sample, the IS peak area
was used to compensate for small deviations in the injec-
tion volume of the GC, to obtain more accurate data.
The IS did not overlap with other chromatographic
peaks present in our samples, as indicated in Table 1.
Unsupervised data analysis: visualizing patterns
in the data set
An unsupervised PCA approach was used to visualize
any potential clustering in the data set, based on the
analysis of a large number of terpenes and cannabinoids.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
207
The data set consisted of 460 samples containing up to
44 identifiable compounds each, which equals over
20.000 variables (i.e., 460 ·44) for evaluation. For the
statistical analysis, each analyzed compound represented
a variable while its quantity (i.e., peak area compensated
for IS) represented the observation. Figure 1a shows an
evaluation of the first two principal components (PC1
and PC2) in the form of a score plot. It is clearly
shown that the main distinction observed in our data
set was between hemp-type samples on the left versus
drug-type samples (Sativa +Indica +hybrids) on the
right. PC1 (28.3%) and PC2 (8.5%) together explained
36.8% of the total variance found in the sample set; almost
twice as high as the model proposed by Hillig
25
(PC1 +
PC2 =19.6%) and equal to the model by Elzinga et al.
24
(PC1 +PC2 =34.6%). Analysis of PC3 added an addi-
tional 7.7% to the total variance explained by our model.
The accompanying loading plot (Fig. 1b) visualizes
howsamplesdifferfromeachother,andwhichcom-
pounds (markers) contribute the most to explaining the
observed grouping. The two components showing the
largest contribution to PC1 are CBD (far left of the plot)
and THC (far right). It is striking that all components, ex-
cept CBD, in the loading plot appear on the right side of
the horizontal (PC1) axis, which means that all are
more prominently present in drug cultivars. Although
the ratio between THC and CBD is a known marker for
the differentiation between hemp- and drug-type canna-
bis,
41
it may be a surprise that not a single terpene
turns out to be a clear marker associated with hemp.
Among the drug-type accessions, a further separation
of samples is seen along the vertical (PC2) axis, separat-
ing samples with relatively higher content of sesqui-
terpenes such as alpha-humulene, beta-caryophyllene,
gamma-cadinene, and eudesma-3,7(11)-diene (upper
half of the plot) versus those typified by relatively higher
content of monoterpenes, such as gamma-terpinene,
alpha-phellandrene, cis-ocimene, alpha-terpineol, and
terpinolene (lower half). Interestingly, cannabinoids
only play a limited role in this distinction, as they are
mainly aligned along the PC1 axis, between the two
groups. The only exception was the minor cannabinoid
CBC, which was more strongly associated with the sub-
group defined by higher sesquiterpene content.
Supervised data analysis: identifying markers
in Sativa versus Indica types
When we applied a supervised (discriminatory) form
of Multivariate Data Analysis, we were able to test
the hypothesis that certain clusters must exist, based
on a-priori knowledge we possess about the sample
set. More specifically, we wanted to verify whether
clear differences exist between the chemical profiles
of Sativa versus Indica samples, assuming these two
classes actually exist. In other words, if indeed mean-
ingful chemical differences exist between the vernacu-
lar Sativa and Indica types, OPLS-DA modeling must
be able to identify the markers that are associated
with the two classes. We therefore included only the
68 accessions labeled as pure Sativa and 63 accessions
labeled as pure Indica, while hemp varieties and hy-
brids were excluded from this particular analysis.
Results of the modeling are visualized in the score
plot shown in Figure 2a. Using the OPLS-DA method,
the horizontal (X) axis visualizes maximum variation
between classes (Sativa versus Indica), while the vertical
(Y) axis shows maximum variation within these classes.
Indeed, the score plots of the OPLS-DA model showed
that the Sativa samples were distinguishable from the
Indica samples with only very mild overlap. Good
model quality was indicated by the values of R2X
(0.387) and Q2 (0.582). The coefficient loading plot
of the OPLS-DA model was used to identify the com-
ponents responsible for sample differentiation on the
score plot (Fig. 2b).
By subsequently constructing an S-plot (figure not
shown) based on the OPLS-DA data, the most signifi-
cant markers associated with our samples could be
identified. We found that the presence of the terpenes
trans-bergamotene, trans-beta-farnesene, delta-3-carene,
and terpinolene was most strongly associated with the
Sativa-type samples, while beta- and gamma-eudesmol,
guaiol, myrcene, and gamma-elemene were the most
prominent markers for the Indica-type samples. Inter-
estingly, we discovered that all hydroxylated terpenes
(marked in Fig. 2b) were most strongly associated
with the Indica type of samples. This begs the question
whether there is a functional correlation between the
presence of the hydroxylated forms of terpenes and
the plant morphology of the Indica type.
Quantitative analysis of Hemp, Sativa, and Indica
All samples labeled as Hemp (n=121), Sativa (n=68),
or Indica (n=63) were quantified with the use of GC
calibration curves. The use of only three standards
(beta-pinene for monoterpenes, alpha-humulene for
sesquiterpenes, and CBD for cannabinoids) for quanti-
fication of all sample components greatly facilitated our
analysis. This approach was based on the observation
that the differences in FID-detector response within
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
208
FIG. 1. (a) Unsupervised PCA score plot of all samples in our sample set (n=460). Solid ellipse indicates 95%
confidence interval based on Hotelling’s T
2
, which can be used to identify outliers. Hemp-type (;n=121) and
drug-type (B;n=339) samples are indicated. (b) PCA loading plot of all samples. Cannabinoids (B),
monoterpenes (-), sesquiterpenes (,). PCA, principal component analysis.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
209
FIG. 2. (a) OPLS-DA score plot discriminating Sativa-labeled (;n=68) from Indica-labeled (B;n=63) samples.
The ellipse in the score plot is a 95% Hotelling’s T
2
, which can be used to identify outliers. (b) OPLS-DA
loading plot. Cannabinoids (B), monoterpenes (-), sesquiterpenes (,). Hydroxylated terpenes are marked with
a dotted box. OPLS-DA, orthogonal projection to latent structure-discriminant analysis.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
210
the chemical class of the monoterpenes, sesquiterpenes,
or cannabinoids, respectively, are relatively small.
27
Because analyzing all 44 sample components by pre-
paring their own standard curves would be an immense
and costly task, one representative compound was cho-
sen to quantify the others in the same chemical class.
Of course, this approach only works if all observed
components have a linear FID response within the an-
alyzed concentration range. Indeed, this was shown to
be true by analyzing selected extracts at different dilu-
tion levels and plotting GC peak area versus dilution
for each observed peak.
The mean calculated content of cannabinoids and
terpenes, reported in mg/g of sample weight, is
shown in Table 2. When comparing cannabinoid pro-
files, all analyzed cannabinoids were found to be pres-
ent in significantly lower quantities in hemp types
compared to Sativa or Indica types ( p<0.05). The
only exception was for CBD, with hemp accessions
(mean: 1.1% CBD) showing a significantly higher
mean content than Sativa or Indica (drug) accessions
(mean: 0.3–0.4% CBD). When comparing Sativa- to
Indica-type samples, we found only significant differ-
ences for the content of CBC and CBG, both of
Table 2. Quantitative Analysis of Cannabinoids, Monoterpenes, and Sesquiterpenes (in mg/g) in accessions labeled
as Hemp, Indica, or Sativa
Mean content (mg/g) –SD
Hemp (n =121) Indica (n =63) Sativa (n =68)
Alpha-2-pinene 0.52 –0.55a 1.21 –1.42b 0.82 –1.40b
Beta-2-pinene 0.19 –0.18a 0.73 –0.55b 0.64 –0.57b
Myrcene 1.06 –0.92a 8.67 –6.95b 2.03 –2.45c
Alpha-phellandrene 0.00 –0.01a 0.01 –0.04ab 0.03 –0.07b
Delta-3-carene 0.01 –0.05a 0.01 –0.06a 0.12 –0.23b
Beta-phellandrene/R-Limonene 0.09 –0.10a 1.38 –1.32b 0.51 –0.37c
Cineol* 0.00 –0.02a 0.00 –0.03a 0.00 –0.00a
Cis-ocimene 0.22 –0.24a 0.80 –0.93b 0.79 –0.91b
Gamma-terpinene 0.00 –0.01a 0.00 –0.00a 0.02 –0.06b
Terpinolene 0.23 –0.32a 0.27 –0.84a 0.94 –1.29b
()Linalool* 0.01 –0.02a 0.31 –0.36b 0.15 –0.12c
Beta-fenchol* 0.00 –0.00a 0.11 –0.17b 0.01 –0.04c
Cis-sabinene hydrate* 0.00 –0.00a 0.08 –0.14b 0.00 –0.03a
Camphor* 0.00 –0.00a 0.00 –0.00a 0.01 –0.06a
Borneol* 0.00 –0.00a 0.01 –0.05b 0.02 –0.14ab
Alpha-terpineol* 0.22 –0.05a 0.41 –0.40b 0.28 –0.14c
Cis-bergamotene 0.21 –0.05a 0.07 –0.11b 0.11 –0.11b
Beta-caryophyllene 0.98 –0.67a 1.94 –1.24b 2.16 –1.57b
Trans-bergamotene 0.04 –0.06a 0.12 –0.14b 0.28 –0.24c
Alpha-guaiene 0.00 –0.02a 0.17 –0.16b 0.29 –0.25c
Aromadendrene 0.00 –0.00a 0.13 –0.19b 0.02 –0.07c
Alpha-humulene 0.27–0.19a 0.61 –0.40b 0.70 –0.61b
Trans-beta-farnesene 0.05 –0.07a 0.31 –0.26b 0.55 –0.37c
Gamma-selinene 0.03 –0.06a 0.26 –0.30b 0.31 –0.33b
Delta-guaiene 0.00 –0.02a 0.03 –0.15b 0.01 –0.03ab
Gamma-cadinene 0.02 –0.05a 0.41 –0.39b 0.36 –0.36b
Eudesma-3,7(11)-diene 0.05 –0.10a 0.56 –0.44b 0.43 –0.43b
Gamma-elemene 0.04 –0.09a 0.64 –0.52b 0.14 –0.26c
Nerolidol* 0.01 –0.03a 0.02 –0.06a 0.00 –0.03a
Trans-beta-caryophyllene 0.06 –0.06a 0.02 –0.05b 0.06 –0.11a
Beta-caryophyllene oxide* 0.00 –0.00a 0.01 –0.05a 0.01 –0.04a
Guaiol* 0.01 –0.03a 0.41 –0.38b 0.02 –0.08a
Gamma-eudesmol* 0.01–0.04a 0.46 –0.36b 0.06 –0.12c
Beta-eudesmol* 0.00 –0.02a 0.25 –0.20b 0.05 –0.10c
Alpha-bisabolol* 0.03 –0.07a 0.36 –0.32b 0.15 –0.17c
THCV 0.14 –0.40a 2.17 –5.26b 0.97 –0.79b
CBD 10.93 –9.47a 2.97 –13.40b 3.77 –15.98b
CBC 0.64 –0.56a 2.06 –1.31b 2.53 –1.35c
CBGM 0.18 –0.17a 0.36 –0.50b 0.73 –4.47ab
THC 3.53 –5.40a 137.10 –53.64b 127.39 –81.21b
CBG 0.28–0.38a 5.09 –4.22b 7.66 –7.19c
Compounds are listed in order of GC elution according to Table 1. Mean values are shown –SD. In each row: values marked with different letters (a,
b, c) are significantly different at p<0.05. Hydroxylated terpenes are marked with an asterisk (*).
SD, standard deviation.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
211
which tend to be somewhat more prominently present
in Sativa types ( p<0.05). The mean THC content of
both types was found to be comparable, with values
of about 127–137 mg/g (or 12.7–13.7%). These values
correspond very well with the average content reported
for Dutch indoor-grown recreational cannabis offered
through coffeeshops.
42
Quantitative analysis of the terpenes confirmed the ob-
servation made by PCA that hydroxylated terpenes
(marked with * in Table 2) are significantly more preva-
lent in Indica than in Sativa accessions. In particular, the
sesquiterpenes guaiol, beta- and gamma-eudesmol, and
alpha-bisabolol were strongly associated with the Indica
type ( p<0.001). Indeed, the presence of such sesquiter-
pene alcohols has previously been reported to be impor-
tant for the chemotaxonomic discrimination of Indica
cannabis varieties originating from Afghanistan.
34
Interestingly, in our study, we found myrcene, on
average the most abundant terpene in our samples,
to be strongly associated with the Indica type of can-
nabis ( p<0.001). In contrast, Hillig
34
measured rela-
tively higher levels of myrcene in Sativa types, while
Lynch et al.
20
found hemp samples to contain signif-
icantly more myrcene than drug-type cannabis. These
conflicting results indicate that selecting single com-
pounds as markers for specific cannabis types may
have its limitations.
Mapping diversity: a practical application
of the chemovar approach
The Dutch medicinal cannabis program currently
provides six different standardized varieties to pa-
tients, with the original one (variety Bedrocan
)hav-
ing been introduced as early as 2003. All six products
were analyzed in the current study, together with an-
other 31 experimental, nonstandardized accessions.
Occasionally, Dutch patients have claimed that the cur-
rent product range does not cover the full spectrum of
varieties needed to treat all medical indications. Such
opinions are commonly based on trial-and-error expe-
riences by individual patients using cannabis obtained
from a variety of sources, including coffeeshops and
other illicit sources, or home-grown cannabis. Although
new varieties have been introduced regularly over the
years (most recently the CBD-rich variety Bedrolite
in 2014), the discussion about the need for new varieties
regularly reappears. A clear and practical chemovar ap-
proach to this discussion may be useful to identify can-
nabis varieties that could be added to the current choice
of products.
In this study, the analysis of a large number of sam-
ples available in The Netherlands has provided us with
a sort of ‘‘chemical diversity map.’’ In Figure 3, the PCA
plot already shown in Figure 1a is repeated, but now
with the six standardized pharmacy products (varieties
Bedrocan, Bedrobinol, Bediol, Bedica, Bedrolite, and
Bedropuur) highlighted to indicate their position on
the map. This visualization makes it clear that two
major sections of the map are not covered, marked as
regions A and B. Based on the sample information
obtained from the original source, we know that region
A corresponds with a cluster of samples popularly la-
beled as Amnesia.
Parallel to this development, a Dutch patient activist
group (Foundation PGMCG, Tilburg, The Nether-
lands) was campaigning for the introduction of a new
cannabis type into the national cannabis program.
Based on their personal experiences with the effects of
various types of cannabis, they independently identified
an Amnesia-type of cannabis to be missing from the
spectrum of currently available varieties. Interestingly,
this group was not aware of this study we were perform-
ing at the same time, which means that our own chem-
ical testing and independent patient experiences both
pointed toward Amnesia as a missing variety, simulta-
neously. This example indicates how real-life experience
with medicinal cannabis can be fully compatible with the
chemical analysis highlighted in our study, indicating
the practical use of the chemovar approach. Together
with the patient group, we were able to identify a canna-
bis variety that displayed the desired chemical profile.
The variety is currently being prepared for standardized
cultivation and is intended to be introduced as a new
medicinal product on prescription later this year.
We are currently working on identifying a cannabis
variety that has the proper chemical content of terpenes
and cannabinoids to match with region B (Fig. 3). Such
a variety would simultaneously need to display the de-
sired botanical characteristics to be cultivated on a large
scale (e.g., yield of flowers per plant, pest resistance,
growth speed, and maximum size).
Conclusion
With modern analytical techniques, the rapid and com-
prehensive analysis of all cannabinoids and terpenes
present in cannabis products has become a feasible op-
tion. This opens up the way to move from a confusing
system of using vernacular cultivar names for the iden-
tification of cannabis products toward a more reliable
and informative system of science-based chemovar
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
212
classification. In our study, we successfully applied this
approach for cannabis differentiation. By analyzing 44
major cannabinoids and terpenes, hemp types of canna-
bis could be efficiently distinguished from drug-type
cannabis based on PCA of the resulting GC data.
Hemp accessions showed a relatively conserved chem-
ical profile characterized by the presence of relatively
more CBD. In contrast, drug accessions showed a
much wider chemical diversity with all cannabinoids
(except CBD) as well as terpenes present in much higher
concentrations than in hemp. An additional distinction
was found between drug accessions characterized by
higher content of monoterpenes, versus a group that
contained significantly higher content of sesquiterpenes.
Potentially, this difference may be caused by evapora-
tion of the relatively more volatile monoterpenes during
drying and storage of some of the analyzed cannabis
accessions.
33
Using a supervised PCA methodology, vernacularly
labeled Sativa and Indica accessions could be well sep-
arated into two distinct groups, which means that true
differences seem to exist in chemical composition be-
tween these two types of cannabis. In contrast to pop-
ular belief, the Sativa and Indica types did not differ in
their average content of the major cannabinoids THC
or CBD. Instead, only CBC and CBG content was
slightly but significantly higher in the Sativa group
(p<0.05). More prominent differences between the
two groups were found in the terpene composition,
specifically in the presence of hydroxylated terpenes as-
sociated with the Indica group. Interestingly, several
other studies have associated the Indica type of canna-
bis with such terpenes, including guaiol and eudes-
mol,
34
fenchol, alpha-terpineol, linalool, geraniol, and
camphor,
24
and linalool.
20
Based on these results, we
believe that hydroxylated terpenes are good chemical
markers for the identification of (vernacular) Indica
cannabis types.
We speculate that the functional link between can-
nabis phenotype and hydroxylated forms of terpenes
may be found in gibberellic acid (GA); a pentacyclic
diterpene acid, and therefore, a hydroxylated terpene
itself. It should be noted that many subtypes of GA
exist. These hormones promote growth and elonga-
tion of cells, and their effect on cannabis development
has been observed by scientists
43–45
as well as cannabis
breeders.
46
Possibly the biosynthetic pathways for GAs
and for hydroxylated terpenes share a common precur-
sor in cannabis. Alternatively, the same catalytic en-
zymes may be involved in their biosynthesis, such as
FIG. 3. Same plot as Figure 1, but indicating currently available Dutch pharmacy products (). Two
regions not covered by currently available products are indicated by dotted circles (A and B).
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
213
a cytochrome P450 monoxygenase
47
or terpene hy-
droxylase.
48
Human selection over centuries of can-
nabis use may have led to a coevolution of plant
morphology (plant shape and size) and terpene profile
(reflected in smell, taste, and medicinal effects) with
GAs as the connecting factor. Further research in live
cannabis plants should be performed to further test
this hypothesis.
As a major limitation, we should note that our study
was based on the analysis of cannabis accessions col-
lected specifically in The Netherlands. The obtained re-
sults may be different when samples from other regions
are analyzed for chemical content. Although in total,
more than 120 terpenes have been identified in various
cannabis products so far,
16
our data suggest that most
of these are only present in trace amounts. The 36 ter-
penes analyzed in our study (Table 1) seem to cover the
whole range of terpenes found in significant quantities,
at least in the samples obtained from our sources in
The Netherlands. Nevertheless, it is possible that sam-
ples from other regions would contain other terpenes
as prominent constituents. As a result, the mapping
of cannabis diversity may be a constantly evolving ef-
fort. New accessions added will make the current
map more complete but could also open up new chem-
ical ‘‘territories’’ that are currently unknown.
Of course, the basis for any evaluation system based
on chemical profiling is that all cannabis constituents
of interest are unambiguously identified. Because of
the large number of terpenes that exist (currently
>20.000 are known) and the large overlap in their
chemical structures and molecular weight, there is a
risk that reported compounds may not be properly
identified. For example, despite earlier reports of their
presence in cannabis samples, the terpenes gamma-
terpineol,
8
phytol,
16,20
carvacrol,
49
and geraniol
24,49
could not be detected in any of our samples. Further-
more, a recent article by Aizpurua-Olaizola et al.
50
used a GC method almost identical to ours, but
reported multiple terpenes that were not found or iden-
tified in our study. To address such discrepancies, there
is a need for a widely accepted and unified method for
terpene analysis in cannabis, to make comparison be-
tween different studies more reliable.
To conclude, it is essential that future cannabis com-
merce allows complete and accurate cannabinoid and
terpene profiles to be available.
51
When the emerging
cannabis industry adopts a common chemical language
across producers, processors, and retailers, the winners
will be the patients. Based on our study, it may be con-
cluded that GC-FID-based metabolomics is a compre-
hensive and effective method to monitor chemical
diversity in cannabis. Achieving a practical, accurate,
and reliable classification system for cannabis, includ-
ing a variety registration system, will require significant
scientific investment and a legal framework that ac-
cepts both licit and illicit forms of this plant. Such a sys-
tem is essential to realize the enormous potential of
cannabis as a multiuse agricultural crop and medicinal
plant. Chemovar mapping may be exactly the tool
needed to achieve this goal.
Acknowledgments
The Trimbos Institute (Utrecht, The Netherlands) is
kindly acknowledged for collecting most of the coffee-
shop samples used in this study, during their annual
evaluation of potency of coffeeshop cannabis (www
.trimbos.nl). We are grateful to the Plant Metab-
olomics Lab (Leiden University, The Netherlands) for
help with the PCA.
Author Disclosure Statement
No competing financial interests exist.
References
1. Kowal MA, Hazekamp A, Grotenhermen F. Review on clinical studies with
cannabis and cannabinoids 2010–2014. Cannabinoids. 2016;11(special
issue):1–18.
2. Hazekamp A, Grotenhermen F. Review on clinical studies with cannabis
and cannabinoids 2005–2009. Cannabinoids. 2010;5:1–21.
3. Erkelens JL, Hazekamp A. That which we call Indica, by any other name
would smell as sweet. Cannabinoids. 2014;9:9–15.
4. Monograph Cannabis flos Version 7.1. Office of Medicinal Cannabis
(OMC): The Hague: The Netherlands, 2014.
5. The Advisability and Feasibility of Developing USP Standards for Medical
Cannabis. Stimuli to the revision process. USP-NF (US Parmacopeia–
National Formulary): Rockville, MD, 2015.
6. Upton R, Craker L, ElSohly M, et al. Cannabis Inflorescence (cannabis
spp.)–standards of identity, analysis, and quality control (revision 2014).
American Herbal Pharmacopoeia (AHP): Soquel, CA, 2014.
7. Small E, Cronquist A. A practical and natural taxonomy for Cannabis.
Taxon. 1976;25:405–435.
8. Hazekamp A, Fischdick JT. Cannabis—from cultivar to chemovar. Drug
Test Anal. 2012;4:660–667.
9. Brodie JS, Di Marzo V, Guy GW. Polypharmacology shakes hands with
complex aetiopathology. Trends Pharmacol Sci. 2015;36:802–821.
10. Szaflarski JP, Bebin EM. Cannabis, cannabidiol, and epilepsy—from re-
ceptors to clinical response. Epilepsy Behav. 2014;41:277–282.
11. Crippa JA, Derenusson GN, Ferrari TB, et al. Neural basis of anxiolytic ef-
fects of cannabidiol (CBD) in generalized social anxiety disorder: a pre-
liminary report. J Psychopharmacol. 2011;25:121–130.
12. Pagano E, Montanaro V, Di Girolamo A, et al. Effect of non-psychotropic
plant-derived cannabinoids on bladder contractility: focus on cannabi-
gerol. Nat Prod Commun. 2015;10:1009–1012.
13. Borrelli F, Pagano E, Romano B, et al. Colon carcinogenesis is inhibited
by the TRPM8 antagonist cannabigerol, a Cannabis-derived non-
psychotropic cannabinoid. Carcinogenesis. 2014;35:2787–2797.
14. Englund A, Atakan Z, Kralj A, et al. The effect of five day dosing with THCV
on THC-induced cognitive, psychological and physiological effects in
healthy male human volunteers: a placebo-controlled, double-blind,
crossover pilot trial. J Psychopharmacol. 2016;30:140–151.
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
214
15. Tudge L, Williams C, Cowen PJ, et al. Neural effects of cannabinoid CB1
neutral antagonist tetrahydrocannabivarin on food reward and aversion
in healthy volunteers. Int J Neuropsychopharmacol. 2014;18:pii: pyu094.
16. Russo EB. Taming THC: potential cannabis synergy and
phytocannabinoid-terpenoid entourage effects. Br J Pharmacol. 2011;
163:1344–1364.
17. Hazekamp A, Fischedick JT, Diez ML, Lubbe A, Ruhaak RL. Chemistry of
Cannabis. In: Mander L, Lui H-W (eds.). Comprehensive Natural Products II
Chemistry and Biology (vol. 3). Elsevier: Oxford, UK, 2010, pp. 1033–1084.
18. Brenneisen R, ElSohly MA. Chromatographic and spectroscopic profiles of
cannabis of different origins: Part I. J Forens Sci. 1988;33:1385–1404.
19. Bakel H van, Stout JM, Cote AG, et al. The draft genome and transcrip-
tome of Cannabis sativa. Genome Biol. 2011;12:R102.
20. Lynch RL, Vergara D, Tittes S, et al. Genomic and Chemical Diversity in
Cannabis. BioRxiv. 2015.
21. Sawler J, Stout JM, Gardner KM, et al. The genetic structure of marijuana
and hemp. PLoS One. 2015;10:e0133292.
22. Pearce DD, Mitsouras K, Irizarry KJ. Discriminating the effects of Cannabis
sativa and Cannabis indica: a web survey of medical cannabis users. J Alt
Comp Med. 2014;20:787–791.
23. Hazekamp A, Pappas G. Self-medication with Cannabis. In: Handbook of
Cannabis. Oxford University Press: Oxford, UK, 2014.
24. Elzinga S, Fischedick J, Podkolinski R, et al. Cannabinoids and terpenes as
chemotaxonomic markers in Cannabis. Nat Prod Chem Res. 2015;3:181.
25. Hillig KW. Genetic evidence for speciation in Cannabis (cannabaceae).
Genet Resour Crop Evol. 2005;52:161–180.
26. McPartland J, Guy G. Correct(ed) vernacular nomenclature of Cannabis.
O’Shaugnessy’s online journal. Edition 4, January 2015.
27. Fischedick JT, Hazekamp A, Erkelens T, et al. Metabolic fingerprinting of
Cannabis sativa L., cannabinoids and terpenoids for chemotaxonomic
and drug standardization purposes. Phytochemistry. 2010;71:2058–2073.
28. Hazekamp A, Simons R, Peltenburg-Looman A, et al. Preparative isolation
of cannabinoids from Cannabis sativa by centrifugal partition chroma-
tography. J Liq Chrom Rel Technol. 2004;27:2421–2439.
29. Hazekamp A, Choi YH, Verpoorte R. Quantitative analysis of cannabinoids
from Cannabis sativa using 1H-NMR. Chem Pharm Bull. 2004;52:718–721.
30. Coffeeshops in Nederland 2014 [Coffeeshops in the Netherlands 2014].
Bureau Intraval: Groningen-Rotterdam, The Netherlands, 2015.
31. Komori T, Nohara T, Hosokawa I, et al. Cannabigerol monomethyl ether, a
new component of hemp. Chem Pharm Bull. 1968;16:1164–1165.
32. Adams RP. Identification of essential oils by ion trap mass spectrometry.
Academic Press Inc.: New York, 1989.
33. Ross SA, ElSohly MA. The volatile oil composition of fresh and air-dried
buds of Cannabis sativa. J Nat Prod. 1996;59:49–51.
34. Hillig KW. A chemotaxonomic analysis of terpenoid variation in cannabis.
Biochem Syst Ecol. 2004;32:875–891.
35. Hillig KW, Mahlberg PG. A chemotaxonomic analysis of cannabinoid
variation in cannabis (cannabaceae). Am J Bot. 2004;91:966–975.
36. Rothschild M, Bergstro
¨m G, Wångberg S. Cannabis sativa: volatile com-
pounds from pollen and entire male and female plants of two variants,
Northern lights and Hawaian indica. Bot J Linnean Soc. 2005;147:387–397.
37. Jolliffe IT, Cadima J. Principal component analysis: a review and recent
developments. Philos Trans A Math Phys Eng Sci. 2016;374:20150202.
38. Eriksson L, Johansson E, Kettaneh-Wold N, et al. Multi- and megavariate
data analysis part 1: basic principles and applications, second edition.
Umetrics Ltd.: Umeå, Sweden, 2006.
39. Trygg J, Wold S. Orthogonal projections to latent structures (O-PLS). J
Chemometr. 2002;16:119–128.
40. Barshan E, Ghodsi A, Azimifar Z, et al. Supervised principal component
analysis: visualization, classification and regression on subspaces and
submanifolds. Pattern Recognit. 2011;44:1357–1371.
41. Brenneisen R, Kessler T. Psychotrope Drogen. V. Die Variabilitat der Can-
nabinoidfu¨ hrung von Cannabispflanzen aus Schweizer Kulturen in
Abha
¨ngigkeit von genetischen und o
¨kologischen Faktoren [Psychotropic
drugs. V. Variability of cannabinoid liberation from Cannabis plants
grown in Switzerland in relation to genetic and ecological factors]. Pharm
Acta Helv. 1987;62:134–139.
42. National Drug Monitor, annual report 2015. Trimbos Institute: Utrecht,
The Netherlands, 2015.
43. Mansouri H, Asrar Z, Amarowicz R. The response of terpenoids to exog-
enous gibberellic acid in Cannabis sativa L. at vegetative stage. Acta
Physiologiae Plantarum. 2011;33:1085–1091.
44. Chailakhyan MK, Khryanin VN. The influence of growth regulators absorbed
by the root on sex expression in hemp plants. Planta. 1978;138:181–184.
45. Heslop-Harrison J, Heslop-Harrison Y. Studies on flowering-plant growth
and organogenesis: IV. Effects of gibberellic acid on flowering and the
secondary sexual difference in stature in Cannabis sativa. Proc Roy Irish
Acad., Sect B. 1960/1961;61:219–232.
46. Rosenthal E, Chong T. Marijuana grower’s handbook: your complete
guide for medical and personal marijuana cultivation. Quick American
Archives: San Francisco, CA, 2010.
47. Janocha S, Schmitz D, Bernhardt R. Terpene hydroxylation with microbial
cytochrome P450 monooxygenases. Adv Biochem Eng Biotechnol. 2015;
148:215–250.
48. Takahashi S, Yeo Y, Greenhagen BT, et al. Metabolic engineering of
sesquiterpe ne metabolism in yeast. Biotechnol Bioeng. 2007;97:
170–181.
49. Turner CE, Elsohly MA, Boeren EG. Constituents of Cannabis sativa L. XVII.
A review of the natural constituents. J Nat Prod. 1980;43:169–234.
50. Aizpurua-Olaizola O, Soydaner U, O
¨ztu¨ rk E, et al. Evolution of the can-
nabinoid and terpene content during the growth of Cannabis sativa
plants from different chemotypes. J Nat Prod. 2016;79:324–331.
51. Piomelli D, Russo EB. The Cannabis sativa versus Cannabis indica debate:
an interview with Ethan Russo, MD. Cannabis and Cannabinoid Research.
2016;1:44–46.
Cite this article as: Hazekamp A, Tejkalova
´K, Papadimitriou S (2016)
Cannabis: from cultivar to chemovar II—a metabolomics approach to
cannabis classification, Cannabis and Cannabinoid Research 1:1,
202–215, DOI: 10.1089/can.2016.0017.
Abbreviations Used
CBC ¼cannabichromene
CBD ¼cannabidiol
CBG ¼cannabigerol
CBN ¼cannabinol
CBGM ¼cannabigerol-monomethylether
FID ¼flame ionization detection
GA ¼gibberellic acid
GC ¼gas chromatography
IS ¼internal standard
MS ¼multiple sclerosis
OPLS-DA ¼orthogonal projection to latent structure-discriminant
analysis
PCA ¼principal component analysis
PCs ¼principal components
THC ¼tetrahydrocannabinol
THCV ¼tetrahydrocannabivarin
Publish in Cannabis and Cannabinoid Research
-Immediate, unrestricted online access
-Rigorous peer review
-Compliance with open access mandates
-Authors retain copyright
-Highly indexed
-Targeted email marketing
liebertpub.com/can
Hazekamp, et al.; Cannabis and Cannabinoid Research 2016, 1.1
http://online.liebertpub.com/doi/10.1089/can.2016.0017
215
