Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis

Abstract

In this paper, we present principal component analysis (PCA) results from a dataset containing 494 cannabis flower samples and 170 concentrate samples analyzed for 31 compounds. A continuum of chemical composition amongst cannabis strains was found instead of distinct chemotypes. Our data shows that some strains are much more reproducible in chemical composition than others. Strains labeled as indica were compared with those labeled as sativa and no evidence was found that these two cultivars are distinctly different chemotypes. PCA of “OG” and “Kush” type strains found that “OG” strains have relatively higher levels of α-terpineol, fenchol, limonene, camphene, terpinolene and linalool where “Kush” samples are characterized mainly by the compounds trans-ocimene, guaiol, β-eudesmol, myrcene and α-pinene. The composition of concentrates and flowers were compared as well. Although the absolute concentration of compounds in concentrates is much higher, the relative composition of compounds between flowers and concentrates is similar.

Full text

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
N
a
t
u
r
a
l
P
r
o
d
u
c
t
s
C
h
e
m
i
s
t
r
y
&
R
e
s
e
a
r
c
h
ISSN: 2329-6836 Natural Products Chemistry & Research
Elzinga et al., Nat Prod Chem Res 2015, 3:4
http://dx.doi.org/10.4172/2329-6836.1000181
Research Article Open Access
Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis
Elzinga S1, Fischedick J2, Podkolinski R1 and Raber JC1*
1The Werc Shop, LLC, Pasadena, CA 91107, USA
2Excelsior Analytical Lab, Inc., Union City, CA 94587, USA
*Corresponding author: Raber JC, The Werc Shop, LLC, Pasadena, CA 91107,
USA, Tel: 855-665-9993; E-mail: jeff@TheWercShop.com
Received June 30, 2015; Accepted July 13, 2015; Published July 20, 2015
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and
Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod Chem Res 3: 181.
doi:10.4172/2329-6836.1000181
Copyright: © 2015 Elzinga S, et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and
source are credited.
Keywords: Cannabis; Tetrahydrocannabinol; Cannabidiol;
Marijuana; Terpenoids; Terpenes; Strains; Taxonomy
Introduction
Cannabis has been used for medicinal and recreational purposes for
millennia. From a taxonomic perspective scientists have been debating
the presence of multiple species for quite some time. Some scientist
delineates 3 species, C. sativa (marijuana type), C. indica (hemp type)
and C. ruderalis (wild type), where others see this as just a variety of 1
species and associated subspecies thereunder [1-4]. is distinction is
exceptionally important as most US States new laws consider Cannabis
sativa L. and subspecies as the only legal cultivars allowed.
It is commonly accepted in cannabis culture to make the distinction
between sativa and indica strains [5]. Indica plantsare said to be short,
densely branched and have wider leaves and are claimed to be sedative
and good for pain relief. Sativa plantsare tall, loosely branched and
have long, narrow leaves. Sativa is claimed to be upliing, stimulating
and recommended for daytime use. is is the typical information
a patient or recreational user will hear when they visit a medicinal
cannabis dispensary or recreational store. However, we only found one
published study that compared indica and sativa strains in patients [6].
is study showed that cannabis was uniformly eective in relieving
symptoms across a wide range of diagnostic categories. Indica strains
appeared superior to sativa strains in improving energy and appetite.
No statistical dierence between sativa and indica was found for pain,
mood, nausea, muscle spasms, seizures, ocular, insomnia, awareness
or neuropathy. Strains were assigned based upon morphology.
is study was not blinded and the observed dierences could be
a result of expectations by the patient.If sativa and indica truly have
dierent physiological eects upon consumption, some compound or
interaction of compounds need to be responsible for this. It has been
shown that cannabidiol (CBD) can inuence the psychoactive eects
of Δ9-tetrahydrocannabinol (THC) [7] and it has been postulated that
the combination of phytocannabinoids and terpenes could result in
complementary or synergistic results oen referred to as the “entourage
eect” [8]. In this paper we use PCA to investigate the analytical results
for cannabinoids and terpenes in 494 cannabis ower samples and
170 cannabis concentrates. is analysis is performed in an attempt
to investigate the potential existence of distinct cannabis chemotypes
that could explain the dierent eects people experience from specic
cannabis strains. Cannabinoids and terpenes were chosen as chemotype
markers as they are considered to be the main physiologically active
constituents in cannabis.Researchers have looked at cannabis from
Abstract
In this paper, we present principal component analysis (PCA) results from a dataset containing 494 cannabis
ower samples and 170 concentrate samples analyzed for 31 compounds. A continuum of chemical composition
amongst cannabis strains was found instead of distinct chemotypes. Our data shows that some strains are much
more reproducible in chemical composition than others. Strains labeled as indica were compared with those labeled
as sativa and no evidence was found that these two cultivars are distinctly different chemotypes. PCA of “OG” and
“Kush” type strains found that “OG” strains have relatively higher levels of α-terpineol, fenchol, limonene, camphene,
terpinolene and linalool where “Kush” samples are characterized mainly by the compounds trans-ocimene, guaiol,
β-eudesmol, myrcene and α-pinene. The composition of concentrates and owers were compared as well. Although
the absolute concentration of compounds in concentrates is much higher, the relative composition of compounds
between owers and concentrates is similar.
a chemotaxonomic perspective as well. Small and Beckstead split C.
sativa L. into three chemotypes based upon CBD/THC ratio [4]. Type
1 has a CBD/THC ratio of <0.5, type II has an intermediate CBD/THC
ratio of 0.5-3.0 and type 3 has a ratio of >3.0. DeMeijer et al showed
that this could be explained genetically by a model involving one locus,
with two alleles. One allele codes for CBDA synthase where the other
codes for THCA synthase. e alleles where shown to be co-dominant
[9]. Pacico et al later showed that classication using just the CBD/
THC values will mask the existence of chemotypes with relatively
high amounts of other cannabinoids [10]. Various authors have tried
using the secondary metabolites in combination with PCA for forensic
investigation of the geographical origin of the plant material [11-
13]. e most elaborate study was performed by El Sohly et al who
analyzed 157 samples from six geographical regions and classied
them using statistical analysis of 175 GC/MS peaks. Although they
managed to dierentiate samples from dierent countries the success
of this approach was limited as not only geographical location but
many other cultivation variables inuenced the chemical composition
of the owers. Much of the cannabis available in the western world
is grown indoors oen with strict control of variables. e use of
controlled lighting cycles, specialized soil, ne-tuned nutrients and
pest control eliminate many of the environmental variables and will
make geographical assigning of the plant dicult if not impossible.
Fischedick et al analyzed 11 cultivars of cannabis for 36 compounds
and managed to discriminate the various cultivars with PCA [14].
Higher levels of cannabinoids correlated positively to higher levels
in terpenoids (R2=0.7688). e authors of this paper showed that it is
possible to grow cannabis with reproducible terpene and cannabinoid
levels over dierent batches as long as environmental conditions and
genetics are standardized. Alterations in grow cycle time, plant stress
and dierent genotype can cause considerable dierences in the
chemical prole.

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 2 of 9
A study by Casano et al investigated the variability of terpene proles
in 16 plants from dierent strains of C. sativa L. [15]. ey separated
the samples into ‘mostly indica’ and ‘mostly sativa’ based upon the
morphological appearance declared by cultivators of the strain. e
studies showed a large variation of relative content of terpenes between
strains and suggest that terpene variation can be used as a tool for
characterization of cannabis bio types. In this study ‘mostly indica’
strains were characterized by dominancy of β-myrcene with limonene
or α-pinene as the second most abundant terpenes. e ‘mostly sativa’
strains were characterized by more complex terpene proles, with some
strains having α-terpinolene or α-pinene as dominant, and some strains
having β-myrcene as dominant with α-terpinolene or trans-β-ocimene
as second most abundant.To our knowledge this is the rst paper
reporting chemo typical dierences using samples that are available to
patients in the California medicinal cannabis dispensaries. Most of the
previous papers use samples collected worldwide and based upon their
reported cannabinoids levels, are not representative of the cannabis
currently available in the United States to patients and recreational
users. It is important to note that doctors oen specically recommend
an indica or sativa strain to their patient, but that the scientic literature
is lacking evidence to support these recommendations. In this paper we
will take a PCA approach to investigate the variation between strains in
the California medicinal marijuana market and also specically look at
the dierences in composition between indica and sativa strains.
Methods
Chemicals
Methanol and water of analytical grade as well as terpene reference
standards were purchased from Sigma Aldrich, St. Louis, MO, USA.
Analytical standards for the cannabinoids were acquired from Restek,
Bellefonte, PA, USA.
Origin of samples
e samples used for this study have been submitted for analysis
to our laboratory by California medicinal marijuana patients in the
period from the beginning of 2012 to the end of 2013. e strain names
for the samples were the names reported by the submitter at the time
of submission.
Statistical analyses
Strains where classied as indica, sativa, hybrid or unknown based
upon the assignment by the cannabis strain database website Leay.
com on June 11, 2015. For concentrates, the whole data set was used
and divided into three categories (high, medium, low CBD). All data
was modulated to express the various compounds as the contribution
to the sum of compounds.
PCA analysis was performed in excel using a macro written by
Tsugawa et al. and is available for free [16] (http://prime.psc.riken.jp/
Metabolomics_Soware/StatisticalAnalysisOnMicrosoExcel/). When
scaling was performed the option “auto scale” was selected.
Analytical measurements
Quantication of THC, CBD, cannabigerol (CBG), Δ9-
tetrahydrocannabinolic acid (THCA), cannabidiolic acid (CBDA)
and cannabigerolic acid (CBGA) was performed using a Shimadzu
prominence UFLC system (Shimadzu Scientic Instruments,
Columbia, MD, USA). e acidic analytical method as published by
Hazekamp et al. was used [17]. Calculation of THCmax, CBDmax
and CBGmax where performed as described in our previous paper
[18]. Terpene content was determined using GC-FID according to the
same approach as Fischedick et al using retention time comparison
with authentic reference, mass spectra, and literature data [14]. A
5% diphenyl and 95% dimethyl polysiloxane column (SHRX5, 15 m,
0.25 mm ID, 0.25 µm lm thickness, Shimadzu Scientic Instruments,
Columbia, MD, USA) and helium carrier gas (Airgas, Radnor, PA,
USA) was used for separation. (Table 1) list the terpenes that were
analyzed in each sample.
Results
Description of ower data set
e dataset contained 494 samples. At least 8 dierent samples
were present for each uniquely identiable strain. A total of 35 dierent
strains where present in the dataset. Table 2 shows the amount of
replicates for each strain and the average, minimum and maximum
THCmax concentration found in the samples. It can be noticed that the
THCmax levels can vary widely even within one strain. In 14 out of 35
strains the dierence between the minimum and maximum level found
diered by more than a factor of 2 and in the highest case (OG Kush)
even more than a factor of 5. is indicates that it will be exceptionally
hard to predict the potency of a ower product based solely upon the
strain name.
Table 2 also shows the assignment to sativa, indica, hybrid or
unknown. irteen (13) strains were assigned as indica, 5 where
assigned as sativa and 14 as hybrid. ree (3) strains did not occur in
the Leay database that was used for the assignment.
Figure 1 shows the distribution of the THCmax (%) in the total
data set. e average THCmax (%) in the data set was 16.8% and the
median for THCmax was 17.1% the distribution was not Gaussian but
showed three peaks. e rst peak is at 4.5-5.49%, which correlates
with low THCmax content in high CBD strains. e second peak is at
15.5-16.49 and the third peak is at 18.5-19.49%. Information regarding
cultivation condition was not present, but it is speculated that these
two peaks represent the averages for outdoor and indoor cultivation
methods.
Figure 2 shows the distribution of the CBDmax (%) in the total data
set. e Figure 2 shows that 478 out of 494 samples had less than 1.49%
CBD indicating that the medicinal marijuana market in California is
dominated by high THC type strains. e average CBDmax (%) in the
data set was 0.6% and the median was 0.3%
e sum of terpenes was plotted against the sum of THCmax and
CBDmax and a R2 value of 0.4248 which is lower than the correlation of
0.77 found by Fischedick [14]. e graph can be found in supplementary
information (Figure 1). is dierence can be explained by the large
variety of strains used in this study where the study by Fischedick was
performed with a limited amount of strains grown and stored under
standardized conditions.For statistical analysis all data was expressed
as a contribution to the sum of all compounds. is modulation of the
data was performed as it is our experience the relative ratios of terpenes
in a strain are more reproducible than the absolute concentration for
a strain. Supplementary information (Figure 2) shows the eect of this
conversion for the 8 main terpenes in 10 dierent Velvet Kush samples.
Absolute data shows more variation than the standardized data. is
eect is likely a result of trichome density [19]. Part of the plant exposed
to more light will have a higher density of trichomes. Also during
trimming of the dried female owers more or less leafy material can be
le behind inuencing the absolute concentration as the trichomes are
the cannabinoid and terpene producing parts of the plant.It was noted
that many owers had a name containing “OG” or “Kush”. Kush is a

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 3 of 9
reference to strains originating from the Hindu Kush region in Central
Asia. e origin of the term “OG” is unknown. To investigate if this
term has any relationship based on the chemical composition of the
plant matter, PCA analysis was performed comparing OG with Kush
type strains. Eleven (11) strains were assigned to the OG group and 5
strains were assigned to the Kush group.
PCA analysis of complete ower data set
e full data set was analyzed without scaling and the scoring and
loading plot can be found in (Figure 3 and 4). PC1 explained 86.3%
of variance and PC2 explained 7.6%. Examination of the loading plot
reveals cannabinoids are responsible for the dierentiation of the
samples. is is expected as the absolute concentration and variation
of cannabinoids is much higher than that of the terpenes therefore
without scaling the cannabinoids will dominate. A grouping of
Harlequin (red), can be noticed. e loading plot indicates that these
strains are dierentiated due to a high CBDmax content. e original
data showed that Harlequin is indeed a high CBD strain and fairly
unique in this aspect. One OG Kush sample (purple) was mixed in with
the Harlequin group and inspection revealed that this indeed was also
Analyzed Terpenes
α-Bisabolol α-Cedrene Limonene α-Phelladerene
Borneol β-Eudesmol Linalool α-Pinene
Camphene (+) Fenchol Menthol β-Pinene
Camphor Geraniol Myrcene Sabinene
Δ3-Carene Guaiol Nerol α-Terpinene
β-Caryophyllene α-Humulene cis-Ocimene α-terpineol
Caryophyllene oxide Isoborneol trans-Ocimene Terpinolene
Table 1: The terpenes analyzed.
Strain Name Replicates Indica/Sativa/Hybrid or Unknown OG or Kush THCmax(%)
Average Min Max
1st Generation Diablo 16 Indica - 20.8 16.2 24.2
Afghan Kush 10 Indica Kush 17.6 14.7 22.0
Alien OG 8 Hybrid OG 19.7 14.6 23.8
Black Mamba 9 Indica - 21.0 19.4 22.8
Blackberry Kush 11 Indica Kush 15.9 12.5 18.0
Blue Dream 31 Hybrid - 16.9 12.2 21.2
Blue Dream Haze 9Hybrid - 17.4 13.6 21.0
Bubba Kush 9 Indica Kush 15.5 10.2 19.4
ChemDawg 14 Hybrid - 16.9 11.2 23.1
Fire OG 23 Hybrid OG 17.3 9.8 20.2
Girl Scout Cookies 19 Hybrid - 15.7 5.8 20.9
Grand Daddy Purple 14 Indica - 16.9 12.2 23.3
Green Crack 16 Sativa - 15.4 11.0 19.1
Harlequin 15 Sativa - 5.0 2.5 12.6
Headband 8Hybrid - 15.5 5.4 22.1
Jack Herer 24 Sativa - 16.9 13.1 21.4
LA Condential 17 Indica - 15.1 8.9 21.7
Larry OG 8 Hybrid OG 17.1 6.2 24.3
Neptune OG 10 Indica OG 18.0 12.6 22.8
NY Sour Diesel 10 Unknown - 15.9 10.8 20.2
OG Herojuana 11 Unknown OG 18.6 15.3 20.7
OG Kush 28 Hybrid - 16.4 4.9 25.0
Platinum OG 8 Indica OG 18.3 15.6 21.4
Pre '98 Bubba 9 Indica - 14.4 9.3 20.3
Purple Kush 8 Indica Kush 13.9 3.8 18.8
SFV OG 12 Hybrid OG 18.9 14.1 23.3
Skywalker 15 Indica - 19.1 9.6 24.5
Skywalker OG 19 Hybrid OG 18.5 13.8 21.7
Sour Diesel 32 Sativa - 16.6 7.7 22.0
Sour OG 10 Hybrid OG 18.0 14.1 24.1
Strawberry Cough 8 Sativa - 15.3 8.7 18.1
Tahoe OG 14 Hybrid OG 17.5 13.2 21.6
Train wreck 16 Hybrid - 14.0 5.9 19.8
True OG 13 Indica OG 18.5 13.4 22.2
Velvet Kush 10 Unknown Kush 21.6 20.0 23.1
Table 2: Summary of the cannabis ower data set.

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 4 of 9
a high CBD sample which was not characteristic of other OG Kush
samples. e same data set was analyzed with scaling. e scoring and
loading plot can be found in (Figure 5 and 6). PC1 explained 22.1% of
variance while PC2 explained 12.5%. Again a grouping of Harlequin
(red) is observed and the loading plot reveals that relatively high
quantities of CBD, α-pinene, guaiol and β-eudesmol are responsible
for the separation in this plot. A second group is observed consisting
of Trainwreck and Jack Herer. e loading plot reveals that this group
is characterized by relatively high amounts of terpinolene, α-terpinene
and α-phellandrene. It is worth noting that the monoterpenoid proles
of these strains were similar to strains described as ‘sativa’ in other
studies [14,15,20]. Plotting of additional principal components did not
reveal any new groupings. As Trainwreck, Jack Herer and Harlequin
were very dominant in the separation, these strains were removed
from the data set to see if additional groups could be detected. When
scaled PCA was performed, PC1 explained 20.5% of the variation,
PC2 explained 10.9%, PC3 8.7% and PC4 7.9%. e PC1 vs PC2 plot
did not show any specic grouping but a general scattered plot of all
the samples. In the PC1 vs PC3 plot a small cluster of Blue Dream could
be noticed (Figure 7). is group showed only slight separation and was
caused by the relatively high concentration of α-pinene in these samples
(Figure 8). When PC1 vs PC4 is plotted a grouping of Green Crack can
be noticed (Figure 9). is group dierentiates itself due to relatively high
amounts of CBGmax, cis-ocimene and trans-ocimene (Figure 10). No
other groupings were noticed in any of the scoring plots.
Comparison of Indica vs Sativa
All unknown and hybrid samples were removed from the data set
and the remaining samples were categorized as either indica (blue) or
sativa (red). Scaled PCA was performed and the scoring and loading
plot can be found in (Figure 11 and 12). PC1 explained 30.2% and PC2
explained 9.2% of the variation. In the scoring plot a mix of indica and
sativa samples can be noticed, but approximately half of the indica
samples separate from the sativa samples. is group is dominant in
the strains 1st Generation Diablo (1stGD), Black Mamba (BM), True
OG (TruenOG), Neptune OG (NepOG) and Skywalker (SkyW). e
loading plot reveals that this group has higher levels of limonene,
0
10
20
30
40
50
60
70
80
0-1.49
1.5-2.49
2.5-3.49
3.5-4.49
4.5-5.49
5.5-6.49
6.5-7.49
7.5-8.49
8.5-9.49
9.5-10.49
10.5-11.49
11.5-12.49
12.5-13.49
13.5-14.49
14.5-15.49
15.5-16.49
16.5-17.49
17.5-18.49
18.5-19.49
19.5-20.49
20.5-21.49
21.5-22.49
22.5-23.49
23.5-24.49
24.5-25.49
25.5-26.49
26.5-27.49
27.5-28.49
28.5-29.49
29.5-30.49
30.5-31.49
Sample count
THCmax (%)
Figure 1: Distribution of THC max (%) in the Flower data set (N=494).
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0-1.49
1.5-2.49
2.5-3.49
3.5-4.49
4.5-5.49
5.5-6.49
6.5-7.49
7.5-8.49
8.5-9.49
9.5-10.49
10.5-11.49
11.5-12.49
12.5-13.49
13.5-14.49
14.5-15.49
15.5-16.49
16.5-17.49
17.5-18.49
18.5-19.49
19.5-20.49
20.5-21.49
21.5-22.49
22.5-23.49
23.5-24.49
24.5-25.49
25.5-26.49
26.5-27.49
27.5-28.49
28.5-29.49
29.5-30.49
30.5-31.49
Sample count
CBDmax (%)
478
Figure 2: Distribution of CBD max (%) in the Flower data set (N=494).
-15
-10
-5
0
5
10
15
20
25
30
-100 -80 -60 -40 -20 0 20
PC2
PC1
Score
Grouping of Harlequin
Figure 3: PCA scoring plot of full cannabis ower data set without scaling.
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide
Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocime ne
Phelladerene
a-Pinene
b-Pinene
Sabinene
a-Terpinene
a-terpineol
Terpinolene
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
PC2
PC1
Loading
Figure 4: PCA loading plot of full cannabis ower data set without scaling.
-10
-8
-6
-4
-2
0
2
4
6
-8 -6 -4 -2 0 2 4 6 8
PC2
PC1
Score
Harlequin
Trainwreck
+
Jack Herer
Figure 5: PCA scoring plot of full cannabis ower data set with scaling.

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 5 of 9
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide
Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocime ne
Phelladerene
a-Pinene
b-Pinene
Sabinene
a-Terpinene
a-terpineol
Terpinolene
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
PC2
PC1
Loading
Figure 6: PCA loading plot of full cannabis ower data set with scaling.
-20
-15
-10
-5
0
5
10
-8 - 6 -4 -2 0 2 4 6 8
PC3
PC1
Score
Blue Dream
Figure 7: PCA (scaled) scoring plot after removing Jack Herer, Train
wreck and Harlequin from the cannabis ower data set PC1 vs PC3.
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide
Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocime ne
Phelladerene
a-Pinene
b-Pinene
Sabinene
a-Terpinene
a-terpineol
Terpinolene
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2
PC3
PC1
Loading
Figure 8: PCA (scaled) loading plot after removing Jack Herer, Train
wreck and Harlequin from the cannabis ower data set PC1 vs PC3.
-20
-15
-10
-5
0
5
-8 - 6 -4 -2 0 2 4 6 8
PC4
PC1
Score
Green Crack
Figure 9: PCA (scaled) scoring plot after removing Jack Herer, Train
wreck and Harlequin from the sample set PC1 vs PC4.
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide
Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool Menthol
Myrcene
Nerol
cis-Ocime ne trans-Ocimene
Phelladerene
a-Pinene
b-Pinene Sabinene
a-Terpinene
a-terpineol
Terpinolene
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2
PC4
PC1
Loading
Figure 10: PCA (scaled) loading plot after removing Jack Herer, Train
wreck and Harlequin from the sample set PC1 vs PC4.
fenchol, α-terpineol, camphene, linalool, THCmax, camphor, geraniol,
β-pinene and ß-caryophyllene.
OG vs Kush
A data set containing only OG type or Kush type sample was
created from the original data (Table 1) and scaled PCA was performed.
e scoring and loading plot can be found in (Figure 13 and 14). PC1
was 29.3%, PC2 was 11.5%. Separation of the two groups can be noticed
although there is an overlapping section. e OG group is dominant
in strains SFV OG (SFV), True OG (True), Tahoe OG (Tahoe), Fire
OG (Fire), Neptune OG (Nept), Larry OG (Larry), Heroijuana OG
(Hero), Platinum OG (Plat) and Skywalker OG (SkyW). is group is
characterized mainly by the compounds α-terpineol, fenchol, limonene,
camphene, terpinolene and linalool. e Kush group is dominant in
strains Velvet Kush (Velv), Blackberry Kush (BB), Purple Kush (Purp).
is group is characterized mainly by the compounds trans-Ocimene,
Guaiol, β-Eudesmol, Myrcene and α-Pinene.
Description of concentrate data set
e concentrate data set consisted of 170 samples. e samples
were assigned to one of three groups. e high CBD group has a
CBDmax/THCmax ratio of >5.0, the medium CBD group had a ratio
of >0.95<5.0 and the low CBD group had a ratio of <0.95. No data was
available regarding the production method of the various concentrates.
e distribution of the THCmax and CBDmax in the concentrate
samples can be found in (Figure 15 and 16). It was noticed that the
frequency of certain THC levels occur more oen. ere appears to be
a peak in frequency at 25-29.9%, 40-44.9, 50-54.9% and 65-69.9%. We
speculate that this is caused by the dierence in production methods
of the concentrates. e concentrates peaking in frequency at 25-

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 6 of 9
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide
Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocim ene
Phelladerene
a-Pinene
b-Pinene
Sabinene
a-Terpinene
a-terpineol
Terpinolene
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
-0.2 -0.1 0 0.1 0.2 0.3 0.4
PC2
PC1
Loading
Figure 12: PCA (scaled) loading plot for indica and sativa.
Figure 11: PCA (scaled) scoring plot for indica (blue) and sativa (red).
29.9% are expected to represent kief products, the peak at 40-44.9% is
expected to represent hash products, the peak at 50-54.9 are expected
to be mainly super critical CO2 extracts and the peak at 65-69.9 are
expected to be hydrocarbon extracts. As with the owers, it can be
noticed that the presence of high amounts of CBD is relatively rare.
Out of 170 samples, 138 samples had less than 4.9% CBD.
PCA analysis of the concentrate data set
Figure 17 shows the PCA scoring plot for the data when scaling
is not applied. PC1 explained 98.7% of the variation, PC2 explained
0.9%. ree distinct groups can be noticed. e loading plot (Figure 18)
shows that CBDmax and THCmax content are the main parameters
that separate the groups. e presence of a low CBD, medium CBD
and high CBD group supports the model of 1 locus with 2 alleles by
DeMeijer et al [9].Scaled PCA was performed as well and the scoring
and loading plot can be found in supplementary information (Figure 3
and 4). PC1 explains 21.4% of the variation and PC2 explains16.2%. No
distinct grouping could be observed in the scoring plots. is analysis

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 7 of 9
Figure 13: PCA (scaled) scoring plot for Kush (red) and OG (blue).
THCmax
CBDmax
CBGmax
a-Bisabolol
Borneol
Camphene
Camphor
3-Carene
b-
Caryophyllene
Caryophyllene
oxide Cedrene
Eudes mol
Fenchol
Geraniol
Guaiol
a-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocim ene
Phelladerene
a-Pinene
b-Pinene
Sabinene
a-Terpinene a-terpine ol
Terpinolene
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
-0.2 -0.1 0 0.1 0.2 0.3 0.4
PC2
PC1
Loading
Figure 14: PCA (scaled) scoring plot for Kush and OG.
was also repeated with just the terpene data, but this did not result
in distinct groups. ree (3) samples separated from the majority of
the samples in the right top eld of the scoring plot. e loading plot
reveals that samples in this region have higher levels of terpinolene,
α-phellanderene, and α-terpinene. It is interesting to note that those
are the same terpenoids that are typical for Jack Herer and Trainwreck
strains. Most likely these concentrates originate from those strains.
PCA analysis of combined ower and concentrate data
To see if the relative chemical composition changes during
processing, PCA was plotted for all the concentrate and ower data. PC1
was 18.6%, PC2 was 13.2%. e scoring and loading plot can be found
in (Figure 19 and 20). No clear grouping of owers vs concentrates
could be noticed but some concentrate samples separate from the
rest of the samples. Inspection of the loading plot and the sample
data revealed these samples have higher relative ratios of β-eudesmol,
guaiol, α-bisabolol and isoborneol. ese terpenes elute relatively late
in GC analyses and it is speculated that these concentrates have been
heated at some time during processing, evaporating part of the more
volatile terpenoids and therefore changing the relative ratio in favor
of the less volatile terpenes. However, the large majority of the owers
and concentrate samples are dispersed among each other in the scoring
plot indicating a similar relative composition of compounds.
Discussion and Conclusions
Cannabis testing labs regularly receive the question “Which strain
has the most THC?”, but as was shown in (Table 1), THCmax content
can be highly variable for a strain. In some cases the level of THCmax
could be 5 times higher in the highest sample compared to the lowest
sample. e data presented shows that strain name cannot be used as
an indication of potency. e observed variation in THC content is
most likely a result of cultivation conditions. is data indicates the

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 8 of 9
0
5
10
15
20
25
30
35
40
Samples counted
Concentration of THCmax (%)
Figure 15: Distribution of THCmax(%) in the concentrate data set (N=170).
0
2
4
6
8
10
12
14
16
18
20
Samples counted
Concentration of CBDmax (%)
138
Figure 16: Distribution of CBDmax(%) in the concentrate data set (N=170).
-25
-20
-15
-10
-5
0
5
10
-120 -100 -80 -60 -40 -20 0 20 40
PC2
PC1
Score
Figure 17: PCA (not scaled) scoring plot for concentrate data.
THC
CBD
CBG
alpha-Bisabolol
Borneol
Camphene
Camphor
delta-3-Caren e
ß-
Caryophyllene
Caryophyllene
oxide
alpha-Cedrene
beta-Eudesmol
(+) Fenchol
Geraniol
Guaiol
α-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocime ne
alpha-
Phelladerene
α-Pinene
ß-Pinene
Sabinene
alpha-
Terpinene
alpha-terpineol
Terpinolene
-0.4
-0.2
0
0.2
0.4
0.6
0.8
-0.8 -0.6 - 0.4 -0.2 0 0.2 0.4 0.6 0.8
PC2
PC1
Loading
Figure 18: PCA (not scaled) loading plot for concentrate data.
Figure 19: PCA (scaled) scoring plot for scaled ower (red) and concentrate
(blue) data.
importance of testing for potency so the consumer knows what to
expect. Strain name is also not a clear indicator of chemical composition.
Variations in chemical composition of samples with the same strain
name describe the importance of broad based chemical proling. By
providing patients with more information regarding the composition
of the cannabis they can rely on chemical composition for reproducible
physiological results instead of strain names who have been shown
to not necessarily correlate with compounds present in the cannabis
owers. PCA of the ower data did not reveal tight clustering of specic
chemotypes but indicates a continuum of varied chemical composition.
In most cases the replicates of a specic strain did not cluster showing
a highly variable chemical composition even within a strain name.
Some strains specically Harlequin, Jack Herer, Trainwreck, Blue
Dream and Green Crack showed much better clustering and seem to
have a more distinct chemical prole than the majority of the strains.
Perhaps these strains are more easily identied by their smell than
other strains as they had relatively high concentrations of specic
terpenes. As our testing perspective here is one from the patient point
of view it would be interesting to investigate chemically distinct strains
for their physiological properties to see to which extent chemotype

Volume 3 • Issue 4 • 1000181
Nat Prod Chem Res
ISSN: 2329-6836 NPCR, an open access journal
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod
Chem Res 3: 181. doi:10.4172/2329-6836.1000181
Page 9 of 9
THC
CBD
CBG
α-Bisabolol
Borneol
Camphene
Camphor
Δ-3-Carene
ß-
Caryophyllene Caryophyllene
oxide
α-Cedrene
ß-Eudesmol
(+) Fenchol
Geraniol
Guaiol
α-Humulene
Isoborneol
Limonene
Linalool
Menthol
Myrcene
Nerol
cis-Ocime ne
trans-Ocime ne
α-Phelladerene
α-Pinene
beta-Pinene
Sabinene
α-Terpinene
α-terpineol
Terpinolene
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
PC2
PC1
Loading
Figure 20: PCA (scaled) loading plot for scaled ower and concentrate data.
might inuence physiological eects. In the comparison of samples
characterized as sativa or indica most of the samples overlapped. No
distinct sativa group clustered independently from indica strains,
however approximately half of the indica samples separated from the
overlapping group. It is likely that many of the strains labeled as sativa
are not in fact pure sativa strains as represented by hemp cultivars (with
relatively higher CBD levels compared to THC) but rather high potency
narrow leaet drug type cultivars with higher THC levels compared to
CBD. is data also indicates another distinct chemotype referred to as
an indica with high levels of limonene, fenchol, α-terpineol, camphene,
linalool, THCmax, camphor geraniol, β-pinene and β-caryophyllene
exists. is chemotype is represented in strains like 1st Generation
Diablo, Black Mamba, Skywalker, Neptune OG and True OG. e
observed data does not support the classication between indica and
sativa as it is commonly presented currently in cannabis culture. A
new classication system is needed to further the medical utility of
cannabis products for patients to enable them to communicate better
with physicians and health care providers.Varieties with the terms OG
and Kush in them are popular in the California medicinal cannabis
market most likely due to their relatively high THC potency and strong
pungent odor. ese strains are also characterized by relatively high
levels of terpenoids with alcohol substitutions. In the comparison of
strains with the term Kush vs OG a dierentiation between the two
categories could be noticed. OG strains had relatively higher levels of
α-terpineol, fenchol, limonene, camphene, terpinolene and linalool
whereas Kush samples were characterized mainly by the compounds
trans-ocimene, guaiol, β-eudesmol, myrcene and α-pinene. Higher
levels of sesquipterpenoid alcohols have been reported to be a potential
distinguishing characteristic of the wide leaet drug type strains
originating from Hindus Kush region of Afghanistan and Pakistan [21].
In media outlets cannabis concentrates are referred to as potentially
more dangerous than herbal cannabis due to higher potency. Although
the average THCmax concentration of concentrates (52.5%) is much
higher than the average concentration in owers (16.8%) the relative
composition is similar. It is therefore expected that physiological eects
should be similar for concentrates and ower if the dose is corrected for
the concentration dierence.
References
1. Beutler JA, Der Marderosian AH (1978) Chemotaxonomy of Cannabis I.
Crossbreeding between Cannabis sativa and C. ruderalis, with analysis of
cannabinoid content. Econ Bot 32: 387-394.
2. Small E, Jui PY, Lefkovitch LP (1976) A numerical taxonomic analysis of
Cannabis with special reference to species delimitation. Sys Botany 1:67-84.
3. Hillig KW, Mahlberg PG (2004) A chemotaxonomic analysis of cannabinoid
variation in Cannabis (Cannabaceae). Am J Bot 91: 966-975.
4. Small E, Beckstead HD (1973) Letter: Cannabinoid phenotypes in Cannabis
sativa. Nature 245: 147-148.
5. Erkelens JL, Hazekamp A (2014) That which we call Indica, by any other name
would smell as sweet. Cannabinoids 9:9-15.
6. Leveroni CV (2001) Differential effects of medicinal marijuana based on strain
and route of administration. Journal of Cannabis Therapeutics 1:43-59.
7. Zuardi AW, Shirakawa I, Finkelfarb E, Karniol IG (1982) Action of cannabidiol
on the anxiety and other effects produced by delta 9-THC in normal subjects.
Psychopharmacology (Berl) 76: 245-250.
8. Russo EB (2011) Taming THC: potential cannabis synergy and
phytocannabinoid-terpenoid entourage effects. Br J Pharmacol 163: 1344-
1364.
9. de Meijer EP, Bagatta M, Carboni A, Crucitti P, Moliterni VM, et al. (2003)
The inheritance of chemical phenotype in Cannabis sativa L. Genetics 163:
335-346.
10. Pacico D, Miselli F, Micheler M, Carboni A, Ranalli P et al (2006) Genetics and
marker-assisted selection of the chemotype in Cannabis sativa L. Molecular
Breeding 17:257-268.
11. ElSohly MA, Stanford DF, Murphy TP (2007) Chemical ngerprinting of
cannabis as a means of source identication. (1st edn) Marijuana and the
cannabinoids. p51-66.
12. Ahmad UK, Muniandy Y, Hassan MS (2014) Physical Analysis and Chemical
Proling of Illicit Herbal Cannabis using Multivariate Analysis. Malaysian
Journal of Forensic Sciences 5: 26-34.
13. Ilias Y, Rudaz S, Mathieu P, Christen P, Veuthey JL (2005) Extraction
and analysis of different Cannabis samples by headspace solid-phase
microextraction combined with gas chromatography-mass spectrometry. J Sep
Sci 28: 2293-2300.
14. Fischedick JT, Hazekamp A, Erkelens T, Choi YH, Verpoorte R (2010)
Metabolic ngerprinting of Cannabis sativa L., cannabinoids and terpenoids
for chemotaxonomic and drug standardization purposes. Phytochemistry 71:
2058-2073.
15. Casano S, Grassi G, Martini V, Michelozzi M (2011) Variations in Terpene
Proles of Different Strains of Cannabis sativa L. ISHS Acta Horticulturae
925: XXVIII International Horticultural Congress on Science and Horticulture
for People (IHC2010): A New Look at Medicinal and Aromatic Plants Seminar
115-121.
16. Tsugawa H, Cajka T, Kind T, Ma Y, Higgins B, et al. (2015) MS-DIAL: data-
independent MS/MS deconvolution for comprehensive metabolome analysis.
Nat Methods 12: 523-526.
17. Hazekamp A, Peltenburg A, Verpoorte R, Giroud C (2005) Chromatographic
and Spectroscopic Data of Cannabinoids from Cannabis sativa L. J Liq
Chromatogr Relat Technol 28: 2361-2382.
18. Elzinga S, Ortiz O, Raber JC (2015) The Conversion and Transfer of
Cannabinoids from Cannabis to Smoke Stream in Cigarettes. Nat Prod Chem
Res 3:163.
19. Happyana N, Agnolet S, Muntendam R, Van Dam A, Schneider B, et al. (2013)
Analysis of cannabinoids in laser-microdissected trichomes of medicinal
Cannabis sativa using LCMS and cryogenic NMR. Phytochemistry 87: 51-59.
20. Hazekamp A, Fischedick JT (2012) Cannabis - from cultivar to chemovar. Drug
Test Anal 4: 660-667.
21. Hillig K (2004) A chemotaxonomic analysis of terpenoids variation in Cannabis.
Biochem Syst Ecol 32: 875-891.
Citation: Elzinga S, Fischedick J, Podkolinski R, Raber JC (2015) Cannabinoids
and Terpenes as Chemotaxonomic Markers in Cannabis. Nat Prod Chem Res 3:
181. doi:10.4172/2329-6836.1000181