Quantitative method for the profiling of the endocannabinoid metabolome by LC-atmospheric pressure chemical ionization-MS

Article abstract of DOI:10.1021/ac0624086

The endocannabinoid system's biological significance continues to grow as novel endocannabinoid metabolites are discovered. Accordingly, a myopic view of the system that focuses solely on one or two endocannabinoids, such as anandamide or 2-arachidonoyl glycerol, is insufficient to describe the biological responses to perturbations of the system. Rather, the endocannabinoid metabolome as a whole must be analyzed. The work described here is based on liquid chromatography coupled with atmospheric pressure chemical ionization mass spectrometry. This method has been validated to quantify, in a single chromatographic run, the levels of 15 known or suspected metabolites of the endocannabinoid system in the rat brain and is applicable to other biological matrixes. We have obtained an endocannabinoid profile specifically for the frontal cortex of the rat brain and have determined anandamide level differences following the administration of the fatty acid amide hydrolase inhibitor AM374.

Full text of DOI:10.1021/ac0624086

Anal. Chem. 2007, 79, 5582-5593
Quantitative Method for the Profiling of the
Endocannabinoid Metabolome by LC-Atmospheric
Pressure Chemical Ionization-MS
John Williams,†,‡ JodiAnne Wood,† Lakshmipathi Pandarinathan,† David A. Karanian,§ Ben A. Bahr,§
Paul Vouros,*,‡ and Alexandros Makriyannis*,†
Center for Drug Discovery, and Department of Chemistry and Chemical Biology, Northeastern University, Boston,
Massachusetts 02115. and Department of Pharmaceutical Sciences, University of Connecticut, Storrs, Connecticut 06269
(2AG), was identified as an agonist to both CB1 and CB2
receptors.5 The endocannabinoids are biosynthesized and biotrans-
formed by a number of enzymes, leading to a variety of enzymatic
products and intermediates, and many of these appear to play a
role in modulating endocannabinoid-related physiological func-
tions. Upon depolarization, induced Ca+2 influx endocannabinoids
are released from postsynaptic neurons and act as retrograde
synaptic messengers on CB receptors located on presynaptic
neurons.6,7 The endocannabinoid system is involved in many
physiological processes, including antinociception, hypothermia,
analgesia, and catalepsy.8
Following the discovery of AEA and 2AG, an effort has been
underway to identify other lipids associated with the endocan-
nabinoid system. For instance, several classes of lipid modulators
found to interact both directly and indirectly with the endocan-
nabinoid system are the N-acyl ethanolamides, acyl glycerols, acyl
amides, and acyl amino acids.9 The “focused metabolomic”
approach has been successful largely due to advances in liquid
chromatography-mass spectrometry (LC-MS) technology.10,11
Recent developments using hybrid quadrupole time-of-flight mass
spectrometers coupled to nanoLC have identified 40 novel acyl
amino acid species with potential bioactivity.12 In the N-acyl
ethanolamide class, palmitoyl ethanolamide (PEA) is a known anti-
inflammatory and antinociceptive agent whose effects are amplified
synergistically when administered with AEA.13 Oleoyl ethanola-
mide (OEA) has been shown to inhibit AEA uptake and degrada-
The endocannabinoid system’s biological significance
continues to grow as novel endocannabinoid metabolites
are discovered. Accordingly, a myopic view of the system
that focuses solely on one or two endocannabinoids, such
as anandamide or 2-arachidonoyl glycerol, is insufficient
to describe the biological responses to perturbations of
the system. Rather, the endocannabinoid metabolome as
a whole must be analyzed. The work described here is
based on liquid chromatography coupled with atmo-
spheric pressure chemical ionization mass spectrometry.
This method has been validated to quantify, in a single
chromatographic run, the levels of 15 known or suspected
metabolites of the endocannabinoid system in the rat
brain and is applicable to other biological matrixes. We
have obtained an endocannabinoid profile specifically
for the frontal cortex of the rat brain and have deter-
mined anandamide level differences following the admin-
istration of the fatty acid amide hydrolase inhibitor
AM374.
The endocannabinoid metabolome can be defined as the full
complement of endogenous lipid signaling mediators that are
associated with the endocannabinoid system. This metabolome’s
composition reflects physiological states (normal and pathological)
known to be associated with the endocannabinoid system. The
endocannabinoid system includes the CB1 receptor, found in high
abundance in the hippocampus, cortex, cerebellum, and basal
ganglia regions of the brain,1 the CB2 receptor, found primarily
in immune tissue,2 and possibly a third cannabinoid receptor,
CB3.3 The first endogenous lipid agonist discovered to have high
affinity to the CB1 receptor was anandamide (AEA), an N-
acylethanolamide derivative of arachidonic acid.4 Soon after, the
monoacyl glyceride of arachidonic acid, 2-arachidonoyl glycerol
(4) Devane, W. A.; Hanus, L.; Breuer, A.; Pertwee, R. G.; Stevenson, L. A.; Griffin,
G.; Gibson, D.; Mandelbaum, A.; Etinger, A.; Mechoulam, R. Science 1992,
258 (5090), 1946-9.
(5) Sugiura, T.; Kondo, S.; Sukagawa, A.; Nakane, S.; Shinoda, A.; Itoh, K.;
Yamashita, A.; Waku, K. Biochem. Biophys. Res. Commun. 1995, 215 (1),
89-97.
(6) Di, Marzo, V.; Melck, D.; Bisogno, T.; De Petrocellis, L. Trends Neurosci.
1998, 21 (12), 521-8.
(7) Piomelli, D.; Beltramo, M.; Giuffrida, A.; Stella, N. Neurobiol. Dis. 1998, 5
(6 Part B), 462-73.
(8) Palmer, S. L.; Thakur, G. A.; Makriyannis, A. Chem. Phys. Lipids 2002, 121
(1-2), 3-19.
(9) Bradshaw, H. B.; Walker, J. M. Br. J. Pharmacol. 2005, 144 (4), 459-65.
(10) Griffiths, W. J. Mass Spectrom. Rev. 2003, 22 (2), 81-152.
(11) Han, X.; Gross, R. W. J. Lipid Res. 2003, 44 (6), 1071-9.
(12) Tan, B.; Bradshaw, H. B.; Rimmerman, N.; Srinivasan, H.; Yu, Y. W.; Krey,
J. F.; Monn, M. F.; Chen, J. S.; Hu, S. S.; Pickens, S. R.; Walker, J. M. Aaps
J. 2006, 8 (3), E461-5.
* To whom correspondence should be addressed. E-mail: a.makriyannis@
neu.edu. Tel.: 617-373-4200. Fax: 617-373-7493. E-mail: p.vouros@neu.edu.
Tel.: 617-373-2840. Fax: 617-373-2693.
† Center for Drug Discovery, Northeastern University.
‡ Department of Chemistry and Chemical Biology, Northeastern University.
§ University of Connecticut.
(1) Tsou, K.; Brown, S.; Sanudo-Pena, M. C.; Mackie, K.; Walker, J. M.
Neuroscience 1998, 83 (2), 393-411.
(2) Munro, S.; Thomas, K. L.; Abu-Shaar, M. Nature 1993, 365 (6441), 61-5.
(3) Breivogel, C. S.; Griffin, G.; Di Marzo, V.; Martin, B. R. Mol. Pharmacol.
2001, 60 (1), 155-63.
(13) Calignano, A.; La Rana, G.; Piomelli, D. Eur. J. Pharmacol. 2001, 419 (2-
3), 191-8.
5582 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
10.1021/ac0624086 CCC: $37.00 © 2007 American Chemical Society
Published on Web 06/29/2007

tion,14 in addition to regulating feeding behavior.15 In the acyl
glycerol class, 2-palmitoyl glycerol (2PG) does not bind directly
to CB2 receptors but has been shown to increase the affinity of
2AG binding when coadministered in equal amounts.16 The
arachidonoyl glycerol ether, also called noladin ether, has CB1
activity and produces physiological responses similar to that of
2AG.17 There are several other lipid modulators with known
bioactivity, such as 2-linoleoyl glycerol and linoleoyl ethanolamide,
that only differ from AEA and 2AG in the size of the acyl chain
and the degree of unsaturation.9
ethanolamide, glyceride, and free acid derivatives of six fatty acid
core structures,. The six fatty acid cores chosen were arachidonic
(AA 20:4n6), docosahexaenoic (DHA 22:6n3), eicosapentaenoic
(EPA 20:5n3), eicosenoic (EA 20:1n9), oleic (OA 18:1n9), and
palmitic (PA 16:0) acids. Combined, the six fatty acids plus their
corresponding glyceride and ethanolamide derivatives total 18
compounds that vary in size and degree of saturation. The choice
of fatty acid cores is based upon evidence that the ethanolamide
and glycerol derivatives are involved in the endocannabinoid
system. In addition to the effects of OEA, PEA, and 2PG on the
endocannabinoid system mentioned previously, docosahexaenoyl
ethanolamide (DHEA) and 2-docosahexaenoyl glycerol (2DHG)
have been identified in the bovine retina and are produced in a
time-dependent manner with AEA.22 In addition, DHEA and
eicosapentaenoyl ethanolamide (EPEA) bind to the CB1 receptor
in rat brains.23 DHA levels in the mouse brain have been shown
to inversely affect the levels of 2AG.24 The role of EA, eicosenoyl
ethanolamide (EEA), and 2-eicosenoyl glycerol (2EG) in the
endocannabinoid system has not been established.
Identifying novel lipid modulators must coincide with improved
methods for their simultaneous quantitation in biological systems
in order to evaluate the physiological changes that occur following
administration of cannabimimetic agents. Initial efforts to quantify
the N-acyl ethanolamides AEA, PEA, and OEA in rat plasma were
performed by GC/MS.18 This method required the time-consum-
ing steps of prepurification and derivatization of the lipid extracts.
Anandamide and 2AG were quantified in CB1 knockout mouse
brain by a reversed-phase LC/MS method using atmospheric
pressure chemical ionization (APCI).19 Switching from GC to LC
eliminated the need for derivatization, but prepurification steps
were still used. Improving upon the original GC/MS method,
Giuffrida introduced the first LC/MS method using electrospray
ionization (ESI) to quantify AEA, PEA, and OEA in rat plasma,
eliminating the prepurification and derivatization steps.20 These
methods all quantify by isotope dilution, using either the molecular
ion [M + H]+ or the sodium adduct [M + Na]+ in selected ion
monitoring (SIM) mode. A more sensitive and selective method
of quantitation than the SIM mode is by selected reaction
monitoring (SRM) mode using a triple-quadrupole mass spec-
trometer, in which only a single fragment ion from a single
precursor is detected. Recently, an SRM-based quantitative LC/
MS/MS method was developed to measure several derivatives
of arachidonic acid, including AEA, 2AG, arachidonoyl glycine,
arachidonoyl dopamine, and arachidonoyl γ-aminobutyric acid, in
male and female rat brains.21 A single internal standard, arachi-
donic acid-d8, was used as the internal standard based upon the
common fatty acid core shared by all analytes. To date, this is
the most comprehensive method available for the quantitative
analysis of the arachidonic core-based endocannabinoids.
The purpose of this work is to develop and validate a robust,
sensitive, and selective LC/MS/MS method for analyzing the
endocannabinoid metabolome in biological matrixes. Here, the
endocannabinoid metabolome has been expanded to include the
EXPERIMENTAL SECTION
Chemicals. Arachidonoylethanolamine and 2-arachidonoyl
glycerol were gifts from the National Institute for Drugs of Abuse.
Arachidonic, oleic, and docosahexaenoic acids were purchased
from Nu-Check Prep (Elysian, MN). Oleoylethanolamide, palmitic
acid, palmitoylethanolamide, eicosapentaenoic acid, ACS grade
acetone, 100 mM phosphate-buffered saline pH 7.4 (PBS), and
HPLC grade ethanol were purchased from Sigma Aldrich (St.
Louis, MO). Arachidonic acid-d8, docosahexaenoic acid-d5, eicose-
noic acid, and eicosapentaenoic acid-d5 were purchased from
Cayman Chemical (Ann Arbor, MI). Palmitic acid-d4 and oleic acid-
d2 was purchased from Cambridge Isotope Laboratories (Andover,
MA). Arachidonoylethanolamide-d4, 2-arachidonoyl glycerol-d5,
oleoylethanolamide-d4, 2-oleoyl glycerol, 2-oleoyl glycerol-d5, palmi-
toylethanolamide-d4, 2-palmitoyl glycerol, 2-palmitoyl glycerol-d5,
docosahexaenoylethanolamide, docosahexaenoylethanolamide-d4,
2-docosahexaenoyl glycerol, 2-docosahexaenoyl glycerol-d5, eicosa-
pentaenoylethanolamine, eicopentaenoylethanolamine-d4, 2-eicosa-
pentaenoyl glycerol, 2-eicosapentaenoyl glycerol-d5, eicosenoic
acid-d2, 2-eicosenoyl glycerol, 2-eicosenoyl glycerol-d5, eicose-
noylethanolamide, and eicosenoylethanolamide-d4 were synthe-
sized in-house. The isotopic purity of each dueterated internal
standard was greater than 98%, except for arachidonic acid-d8,
which was 80%. There was no contribution from the dueterated
standards to the unlabeled analytes arising from incomplete (d0)
deuterium incorporation. HPLC grade water, methanol, and
chloroform were purchased from Fisher Scientific (Pittsburgh,
PA). Fatty acid free bovine serum albumin (BSA) was purchased
from EMD Biosciences (San Diego, CA). Rat brain frontal cortex
sections were provided by Dr. Ben Bahr at the University of
Connecticut, Storrs.
(14) Rakhshan, F.; Day, T. A.; Blakely, R. D.; Barker, E. L. J. Pharmacol. Exp.
Ther. 2000, 292 (3), 960-7.
(15) Rodriguez De Fonseca, F.; Gorriti, M. A.; Bilbao, A.; Escuredo, L.; Garcia-
Segura, L. M.; Piomelli, D.; Navarro, M. Neurotoxic. Res. 2001, 3 (1), 23-
35.
(16) Hanus, L.; Gopher, A.; Almog, S.; Mechoulam, R. J. Med. Chem. 1993, 36
(20), 3032-4.
(17) Hanus, L.; Abu-Lafi, S.; Fride, E.; Breuer, A.; Vogel, Z.; Shalev, D. E.;
Kustanovich, I.; Mechoulam, R. Proc. Natl. Acad. Sci. U.S.A. 2001, 98 (7),
3662-5.
General Method of Synthesis: N-Acylethanolamides and
Monoglycerides. 1-(3-Dimethylaminopropyl)-3-ethylcarbodiimide
(18) Giuffrida, A.; Piomelli, D. FEBS Lett. 1998, 422 (3), 373-6.
(19) Di, Marzo, V.; Breivogel, C. S.; Tao, Q.; Bridgen, D. T.; Razdan, R. K.;
Zimmer, A. M.; Zimmer, A.; Martin, B. R. J. Neurochem. 2000, 75 (6), 2434-
44.
(22) Bisogno, T.; Delton-Vandenbroucke, I.; Milone, A.; Lagarde, M.; Di, Marzo,
V. Arch. Biochem. Biophys. 1999, 370 (2), 300-7.
(20) Giuffrida, A.; Rodriguez de Fonseca, F.; Piomelli, D. Anal. Biochem. 2000,
280, (1), 87-93.
(23) Sheskin, T.; Hanus, L.; Slager, J.; Vogel, Z.; Mechoulam, R. J. Med. Chem.
1997, 40 (5), 659-67.
(21) Bradshaw, H. B.; Rimmerman, N.; Krey, J. F.; Walker, J. M. Am. J. Physiol.
Regul. Integr. Comp. Physiol. 2006, 291, (2), R349-58.
(24) Watanabe, S.; Doshi, M.; Hamazaki, T. Prostaglandins, Leukotrienes Essent.
Fatty Acids 2003, 69 (1), 51-9.
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5583

hydrochloride (0.6 mmol) was added to a solution of fatty acid
(0.5 mmol), ethanolamine, or glycerol (0.75 mmol), and N,N-
dimethylaminopyridine (0.5 mmol) in 20 mL of acetonitrile at
0 °C. The reaction mixture was warmed to room temperature and
stirred for 2-8 h. After the reaction was completed, as determined
by TLC analysis (eluent, 30% acetone in hexane; stain developing
agent, 5% phosphomolybdic acid in 95% ethanol), stirring was
stopped and the solvent was evaporated under vacuum. The crude
residue was loaded on SiO2 and purified by the Biotage SP1 flash
chromatography system (gradient method with acetone and
hexane from 5 to 60%). Pure fractions were pooled, and the solvent
was evaporated under vacuum to provide pure N-acylethanola-
mide/monoglyceride (75-90% yield) that was analyzed by NMR
to ensure characteristic signals consistent with the assigned
structures. The NMR assignment data for each endocannabinoid
metabolite synthesized in-house can be found in the Supporting
Information.
Preparation of Standards. Dried, purified samples of the 18
analytes and 18 deuterated analogues, ranging in weight from 1
to 5 mg, were diluted with an appropriate volume of ethanol to
prepare separate 1 mg/mL stock solutions. The 200-µL aliquots
of each individual analyte stock solution were combined in a 15-
mL Falcon tube (Beckton Dickson, Franklin Lakes, NJ) to form
a 3.6-mL stock solution comprising all analytes and their labeled
analogues. The 90-µL aliquots of the combined stock solution were
transferred to 40 individual 0.6-mL siliconized microcentrifuge
tubes (Fisher Scientific). Each microcentrifuge tube contained
5000 ng of each analyte. The procedure was repeated with the 18
deuterated analogue stock solutions to generate 40 microcentri-
fuge tubes containing 5000 ng of each deuterated analog. The
combined stock solution in each tube was evaporated to dryness
under nitrogen at 30 °C (N-Evap 111, Organomation Associates,
Berlin, MA). Upon dryness, each individual tube was capped,
sealed with Parafilm (American National Can, Neenah WI), and
stored at -80 °C until used for preparing calibration curves. A 20
mg/mL solution of fatty acid-free BSA was prepared by diluting
400 mg of BSA with 20 mL of water. The BSA solution was kept
at 4 °C until analysis.
Standard Curve Preparation. On the day of analysis, a single
vial of the dried analyte stock and a single vial of the dried
deuterated analogue stock were reconstituted with 500 µL of
ethanol each to make 10 ng/µL working solutions. The 100 µL of
working analyte solution was diluted with 900 µL of BSA in a 1.5-
mL siliconized centrifuge tube (Fisher Scientific) to make a 1 ng/
µL solution. A 100 µLof the 1 ng/µL solution was further diluted
with 900 µL of BSA to make a 100 pg/µL solution of the combined
analyte standard. A 1 ng/µL solution in BSA of the combined
deuterated analogues was prepared in a manner identical to that
for the 1 ng/µL analyte solution. Solutions of 1 ng/µL of both the
combined analyte standard and the combined deuterated analogue
standard were also prepared in ethanol, by a 1:10 dilution in
ethanol of the 10 ng/µL working solutions. Aliquots of the 100
pg/µL and the 1 ng/µL BSA solutions were diluted with BSA
solution to generate calibration standards of 0, 5, 10, 25, 50, 100,
250, 450, and 500 pg/µL. Twenty microliters of the 1 ng/µL
deuterated analogue solution in BSA was added to each calibration
standard to act as internal standard (IS). The total volume of each
calibration standard was 200 µL and contained 100 pg/µL of IS.
One extraction blank of 200 µL of 20 mg/mL BSA solution was
made. Quality control (QC) samples in BSA were made at analyte
concentrations of 7, 21, 120, and 400 pg/µL and spiked with 20
µL of 1 ng/µL IS in BSA. Three reference samples containing
300 pg/µL each of analyte and IS were prepared using the 1 ng/
µL solutions in ethanol. Three reference extraction samples of
the same concentration were also prepared using the 1 ng/µL
solutions in BSA. QC and sample concentrations were calculated
from the calibration curve using an equal weight linear regression
(X) algorithm, where X represents concentration. The calibration
curves were constructed from the ratios of the peak areas of the
analytes versus the IS.
Standard and Sample Extraction. The extraction procedure
for the calibration standards, QC, reference, and brain samples
is modified from the Folch extraction.25 The 50 µL of ice cold PBS
and 400 µL of acetone were added to each calibration standard,
blank extraction sample, and reference extraction sample, and then
each mixture was vortexed for 10 s to precipitate protein. The
samples were then centrifuged at 20800g for 5 min (Eppendorf
Centrifuge 5417C, Hamburg Germany). The supernatant was
transferred to a new 1.5mL centrifuge tube and dried under
nitrogen until the acetone was removed. A total of 100 µL of PBS,
250 µL of methanol, and 500 µL of chloroform were added to the
remaining supernatant, which was vortexed for 10 s, and then
centrifuged at 20800g for 5 min to separate the two phases. The
lower organic phase was transferred to a new 1.5-mL centrifuge
tube and evaporated to dryness under nitrogen. The dried
extracted standards were reconstituted in 200 µL of ethanol,
vortexed for 15 s, sonicated for 15 s, and then centrifuged at
20800g for 5 min before transferring them to HPLC vials for
analysis.
Brain samples were received on dry ice and immediately stored
at -80 °C until the following day when they were processed and
analyzed. The frozen brain sections were weighed prior to
homogenization in ice cold acetone/PBS, pH 7.4 (3:1) and 5 µL
of 1 ng/µL IS, followed by centrifugation at 10000g for 10 min at
4 °C. The resulting supernatant was dried under nitrogen until
the acetone was removed. To the remaining supernatant, 100 µL
of PBS, one volume of methanol, and two volumes of chloroform
were added for liquid-liquid-phase extraction of the lipids. The
two phases were separated by centrifugation, and the bottom
organic layer was evaporated to dryness under nitrogen. Brain
extracts were reconstituted in 50 µL of ethanol, vortexed, sonicated
briefly, and centrifuged at 20800g for 5 min prior to analysis.
In addition to the reconstituted sample, serial dilutions of 10×
and 100× in ethanol prepared from each sample were also
analyzed.
LC-MS Analysis. The liquid chromatography system used
for analysis was an Agilent 1100 HPLC (Agilent Technologies,
Wilmington DE). The autosampler was kept at 4 °C. Liquid
chromatography was performed on a Zorbax SB-CN 2.1 × 50 mm,
5-µm, 80-Å column (Agilent Technologies). Mobile phase A
consisted of 10 mM ammonium acetate, adjusted to pH 7.3 using
ammonium hydroxide. Mobile phase B consisted of 100% metha-
nol. Gradient elution was performed at a flow rate of 500 µL/min.
The initial composition of the gradient was 10% B. This was held
(25) Folch, J.; Lees, M.; Sloane Stanley, G. H. J. Biol. Chem. 1957, 226 (1),
497-509.
5584 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Figure 1. Structures of the endocannabinoid metabolites and their internal standards. Each analyte’s SRM transition is listed in the last row
of text. The specific collision energy used is shown in brackets.
isocratically for the first minute. At 1 min, mobile phase B
increased to 60% B within 0.5 min. From 1.5 to 10 min, the gradient
ramped linearly from 60% B to 75% B. At 10 min, the gradient
increased to 95% B within 0.5 min, where it was held for 1.5 min.
At 12.5 min, the gradient returned to initial conditions of 10% B
to re-equilibrate the column. The total run time was 15 min.
Chromatography was optimized to separate the free fatty acids
during the first 6 min of the run and the ethanolamide and
glyceride derivatives during the latter 9 min.
psi. During the last 9 min of the LC run, the Quantum operated
in positive mode with a probe temperature of 350 °C, a source
temperature of 250 °C, corona discharge current of 6 µA, and a
sheath gas setting of 30 psi. For both ionization modes, the
collision pressure was kept at 1 mTorr and centroid data were
acquired. The resolution setting for Q1 and Q3 was 0.2, and each
SRM scan time was 0.1 s. The SRM transitions that were
monitored for each analyte and deuterated analogue are listed in
Figure 1.
The column was connected to a TSQ Quantum Ultra-triple-
quadrupole mass spectrometer (Thermo Fisher, San Jose CA).
Analytes were ionized via APCI. The instrument was operated in
both positive and negative ionization modes. During the first 6
min of the LC run, the Quantum operated in negative mode with
a probe temperature of 380 °C, a source temperature of 250 °C,
corona discharge current of 15 µA, and a sheath gas setting of 30
Validation Procedure. A validation program based on the
FDA guidelines for linearity, inter- and intraday accuracy and
precision, recovery, and stability was executed for the quantitation
of the previously mentioned lipid modulators comprising the
endocannabinoid metabolome.26
Linearity. Eight nonzero calibration standards, with analyte
concentrations ranging from 5 to 500 pg/µL were prepared and
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5585

analyzed in triplicate in three separate analytical runs. The
calibration curves were calculated by least-squares linear regres-
sion using an equal weighting factor. The standard concentrations
were back-calculated from constructed calibration curves for each
analyte. The deviation from nominal concentrations over three
runs should be within (15% ((20% at the lower limit of quanti-
tation) for the calibration standards.
Accuracy and Precision. The accuracy and precision of the
assay were established by analyzing QC samples of each analyte.
Three replicates of each sample were analyzed together with a
complete set of calibration standards in three analytical runs.
The intra-assay accuracy was determined as the percent dif-
ference between the mean concentration per analytical run and
the nominal concentration. The interassay accuracy was deter-
mined as the percent difference between the mean concentra-
tion after three analytical runs and the nominal concentration.
The coefficient of variation provided the measure of intra- and
interassay precision. The accuracy and precision of the mea-
sured concentrations varied less than 15% from the nominal
concentrations.
Table 1. Sample Recovery Following Precipitation and
Extraction Procedures
reference sample
extraction
analyte
nonextracted
extracted
recovery, %
AA
95.9
103.6
95.5
86.8
104.3
96.0
90.5
100.7
100.5
97.8
97.8
97.7
99.9
101.4
96.4
98.6
96.2
94.9
102.7
97.2
93.2
DHA
EPA
AEA
PEA
OEA
DHEA
EPEA
EEA
2AG
97.2
95.1
102.3
103.9
101.6
118.5
100.1
88.1
100.1
101.5
101.5
120.2
96.5
86.9
2PG
89.2
92.7
2OG
2DHG
2EPG
2EG
93.7
88.9
89.8
92.2
95.0
92.3
85.0
79.2
Table 2. Extraction Recovery (Mean ( SEM) of AEA,
2-AG, and EPA from Rat Brain Homogenate As
Determined by the Standard Additions Method
Stability. The stability of each analyte was investigated during
all steps of the analysis, including the stability in the stock
preparations in both ethanol and BSA over a 2-month period.
Freeze-thaw stability of each analyte was also determined over
four cycles at -80 °C. Analytes were considered stable in the
biological matrix when 85-115% of the initial concentration was
found. In stock solutions, analytes were considered stable when
90-110% of the initial concentration was found.
regression
extraction
R2 value
recovery, %
AEA
2-AG
EPA
0.996
0.996
0.959
113.9 ( 24.1
105.7 ( 22.0
96.4 ( 17.1
this method requires a mobile-phase system that would promote
positive ionization of the ethanolamide and glycerol derivatives
and negative ionization of the free acids by APCI conditions. In
order to determine the effect of pH on ionization in both modes,
several different additives, such as formic acid, acetic acid,
ammonium hydroxide, methyl morpholine, and ammonium ac-
etate, were studied. The type of organic solvent, either acetonitrile
or methanol, was also tested to determine the best choice for
ionization and elution strength. The best results were obtained
using a 10 mM ammonium acetate pH 7.3 aqueous solution as
solvent A and pure methanol as solvent B. Second, a chromato-
graphic separation was required that could isolate the free acid
metabolites from the ethanolamide and glyceride metabolites so
that the ionization mode could be switched from negative to
positive to ionize each component. Due to the hydrophobic
character of all endocannabinoid metabolites, the retentiveness
of typical C18 and C8 reversed-phase columns was too great to
provide adequate separation. However, the cyano column, with
its low retention and alternative selectivity, was able to separate
the components with enough resolution to operate the method
in both modes successfully.
With a suitable solvent system and chromatographic separation
established, precursor and product ions for each analyte were
determined by direct infusion. The most intense product ion was
monitored for each analyte. Choosing the best product ion
fragment for each SRM transition was complicated because the
core fatty acid tails tended to fragment into many different
products. However, characteristic losses from each of the deriva-
tive types were exploited and suitable product ions were deter-
mined. For the ethanolamide derivatives, fragmentation of the
amide bond between the fatty acid core and the ethanolamine
Recovery. The efficiency of the sample extraction procedure
was determined by comparing the calculated concentrations of
the nonextracted reference samples to the extracted reference
samples. The percent recovery was calculated from the extracted/
nonextracted ratio.
To support the use of BSA as the matrix for the preparation
of standards, the extraction efficiency was also determined from
rat brain homogenate by the method of standard additions. Briefly,
several rat brains were homogenized following the exact proce-
dure as described in the Standard and Sample Extraction section.
The pooled homogenate was divided into three sets of five 1-mL
aliquots. In each set, aliquots were spiked with 0, 2.5, 5.0, 12.5,
and 25 ng of AEA, 2-AG, and EPA standards. Following sample
extraction and reconstitution, the three sets were analyzed using
the same LC-MS conditions described previously. The peak areas
of each component were used to create the standard addition
plot, from which the levels of each component were determined
for each spiked aliquot. Subtraction of the endogenous level of
each analyte from the spiked sample levels yielded the amount
of spiked standard as determined by the standard additions
method. The ratio of the measured spike amount to the true spike
amount was used to determine the efficiency of the sample
extraction.
RESULTS AND DISCUSSION
Method Development. The structural diversity between the
members of the endocannabinoid metabolome analyzed by the
current method posed two main challenges to its success. First,
(26) Food and Drug Administration (FDA) Guidance for Industry-Bioanalyti-
cal Method Validation, 2001; http://www.fda.gov/guidance/4252fnl.htm
(accessed June 2007).
5586 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Table 3. Inter- and Intraday Accuracy and Precision of the QC Sample at Four Concentrationsa
QC 1: 7 pg/µL
QC 2: 21 pg/µL
QC 3: 120 pg/µL
QC 4: 400 pg/µL
accuracy, precision,
accuracy, precision,
accuracy, precision,
accuracy, precision,
analyte
AEA
mean
intraday 1 6.6
%
%
mean
%
%
mean
%
%
mean
%
%
6
11.8
3.1
20.9
20.3
21.0
20.7
0.4
3.2
0.1
1.2
3.2
3.8
3.5
3.5
115.5
111.8
123.0
116.8
3.8
6.9
2.5
4.4
3.1
1.5
1.5
2.0
392.0
388.4
408.8
396.4
2.0
2.9
2.2
2.4
5.1
4.2
3.2
4.2
intraday 2 6.8
intraday 3 7.1
2.5
1.1
3.2
16.3
10.4
interday
6.8
PEA
OEA
intraday 1 6.8
intraday 2 7.2
intraday 3 6.9
3.2
3.5
1
5.6
2.9
9.1
5.9
21.3
20.3
21.8
21.1
1.5
3.5
3.7
2.9
3.3
8.6
2.0
4.6
119.2
114.5
120.3
118.0
0.6
4.6
0.3
1.8
7.0
3.8
2.1
4.3
390.4
400.6
403.5
398.2
2.4
0.1
0.1
0.9
5.6
6.0
1.1
4.2
interday
7.0
2.6
intraday 1 6.8
intraday 2 6.9
intraday 3 7.4
3.3
2
5.8
3.7
3.7
1.7
7.9
4.4
20.6
21.4
20.1
20.7
2.0
2.0
4.1
2.7
2.0
3.2
6.0
3.7
124.8
115.8
121.4
120.7
4.0
3.5
1.1
2.9
7.0
2.4
0.4
3.3
388.7
407.8
390.4
395.6
2.8
2.0
2.4
2.4
2.6
3.0
3.4
3.0
interday
7.0
DHEA intraday 1 6.7
intraday 2 6.7
4.6
3.7
0.5
2.9
7.9
9.3
6.6
7.9
21.9
21.2
21.0
21.4
4.5
1.2
0.2
2.0
3.9
5.5
4.7
4.7
125.9
110.3
120.4
118.9
4.9
8.1
0.4
4.5
2.9
1.2
0.9
1.7
410.1
390.2
392.9
397.7
2.5
2.5
1.8
2.3
4.8
1.0
1.4
2.4
intraday 3 7.0
interday
6.8
EPEA intraday 1 6.8
intraday 2 6.0
3.1
14.7
7.0
11.2
10.6
7.4
19.0
19.1
19.8
19.3
5.0
4.3
1.1
3.4
8.2
13.3
5.4
134.6
123.2
130.3
129.4
12.1
2.7
5.4
3.2
2.0
3.5
431.6
424.5
423.1
426.4
7.9
6.1
5.8
6.6
2.9
6.6
2.1
3.9
intraday 3 7.5
8.6
interday
6.7
8.3
9.7
9.0
7.8
EEA
2AG
2PG
2OG
intraday 1 6.9
intraday 2 6.5
intraday 3 7.1
1.3
7.7
1.5
3.5
4.3
7.5
2.6
4.8
20.9
20.9
20.8
20.9
0.4
0.4
0.9
0.5
3.7
1.0
3.4
2.7
120.2
116.3
123.2
119.9
0.1
3.1
2.7
2.0
5.3
3.1
3.1
3.8
407.3
401.0
397.4
401.9
1.8
0.2
0.7
0.9
9.0
6.0
2.0
5.6
interday
6.8
intraday 1 6.6
intraday 2 6.9
intraday 3 5.8
6.1
1.0
17.7
8.3
12.8
13.1
7.9
21.8
17.9
18.4
19.4
3.8
14.8
12.2
10.3
6.5
10.2
4.3
121.6
111.5
119.4
117.5
1.3
7.1
0.5
3.0
0.2
13.8
11.7
8.6
376.2
428.8
380.4
395.1
6.0
7.2
4.9
6.0
4.3
12.9
5.0
interday
6.4
11.3
7.0
7.4
intraday 1 6.3
intraday 2 6.8
intraday 3 5.7
9.8
2.9
18.5
10.4
8.2
6.7
12.3
9.1
20.5
22.8
21.7
21.7
2.3
8.6
8.1
6.3
6.3
5.8
9.4
7.2
116.4
121.8
107.4
115.2
3.0
1.5
10.5
5.0
3.2
5.4
8.2
5.6
408.4
415.6
418.3
414.1
2.1
3.9
4.6
3.5
2.3
2.6
7.4
4.1
interday
6.3
intraday 1 7.2
intraday 2 6.9
intraday 3 7.4
3.4
0.8
6.1
3.4
5.1
13.3
8.1
20.5
19.1
22.2
20.6
2.5
8.9
5.7
5.7
7.5
2.3
9.9
6.6
122.6
122.8
130.7
125.4
2.1
2.3
8.9
4.4
11.1
4.7
13.2
9.7
390.4
428.5
408.8
409.2
2.4
7.1
2.2
3.9
10.2
1.5
2.8
4.8
interday
7.2
8.8
2DHG intraday 1 6.6
intraday 2 7.2
6.0
2.3
10.4
6.2
7.9
14.7
12.4
11.7
23.6
22.3
21.3
22.4
12.5
6.3
11.3
3.6
127.4
117.4
117.6
120.8
6.2
2.1
2.0
3.4
2.2
0.8
4.0
2.3
368.2
394.5
399.5
387.4
7.9
1.3
0.1
3.1
0.6
4.5
5.0
3.4
intraday 3 6.3
1.6
5.2
interday
6.7
6.8
6.7
2EPG intraday 1 7.4
intraday 2 6.6
5.5
5.5
4.3
5.1
5.6
13.5
13.9
11.0
23.3
18.0
20.1
20.5
11.8
14.1
4.4
9.1
7.5
8.2
8.3
119.7
119.9
124.5
121.4
0.3
0.1
3.7
1.4
9.4
7.8
8.6
8.6
383.4
443.2
405.7
410.8
4.2
10.8
1.4
7.2
14.8
11.6
11.2
intraday 3 7.3
interday
7.1
10.1
5.5
2EG
AA
intraday 1 6.9
intraday 2 7.3
intraday 3 7.3
1.6
4.2
4.6
3.5
8.5
12.9
8.3
22.1
23.2
18.7
21.3
5.0
10.4
10.7
8.7
3.9
8.0
12.7
8.2
108.7
121.5
116.5
115.6
9.5
1.2
2.9
4.5
2.6
7.5
10.4
6.9
392.0
405.5
424.3
407.3
2.0
1.4
6.1
3.2
2.0
10.0
10.7
7.6
interday
7.2
9.9
intraday 1 5.9
intraday 2 6.6
intraday 3 5.8
16.1
5.3
16.6
12.7
1.3
1.9
14.7
6.0
18.6
18.4
21.4
19.5
11.6
12.2
1.8
7.3
2.0
12.7
7.3
110.5
111.3
123.1
115.0
7.9
7.2
2.5
5.9
6.4
8.7
1.3
5.5
392.7
386.9
396.1
391.9
1.8
3.2
1.0
2.0
7.1
5.7
7.3
6.7
interday
6.1
8.5
DHA
EPA
intraday 1 7.7
intraday 2 5.8
intraday 3 8.1
10.4
17.6
16.4
14.8
13.2
19.7
8.0
21.1
20.9
21.1
21.0
0.5
0.3
0.5
0.4
6.9
10.6
17.1
11.5
119.5
121.6
118.0
119.7
4.1
1.3
1.6
2.3
9.8
4.2
0.8
4.9
384.5
410.8
420.1
405.1
3.9
2.7
5.0
3.9
5.6
2.8
4.6
4.3
interday
7.2
13.6
intraday 1 6.7
intraday 2 6.10
intraday 3 6.80
4.3
12.3
2.4
1.2
6.1
13.3
6.9
20.3
20.1
20.3
20.2
3.1
4.4
3.5
3.7
8.9
8.1
3.8
6.9
108.0
119.3
122.5
116.6
10.0
0.6
2.1
4.2
2.7
2.6
9.3
4.9
417.8
401.3
386.1
401.7
4.4
0.3
3.5
2.7
2.5
5.7
2.7
3.6
interday
6.5
6.3
a Accuracy % represents the difference between the measured and true value. Precision % represents the average standard deviation of the
measurements (n ) 3).
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5587

Figure 2. Extracted MRM chromatograms of the endocannabinoid metabolites in the reference standard, 500 pg of each analyte on column.
yielded protonated ethanolamine products of m/z 62. Internal
standards for these metabolites are labeled with four deuteriums
on the ethanolamine tail, resulting in a product ion of m/z 66.2.
In a fashion similar to the ethanolamides, the glyceride derivatives
fragmented at the ester linkage between the fatty acid and the
glyceride moiety. However, in these cases, the charge was
typically retained on the fatty acid core, giving product ions that
were 91 amu less than the precursor m/z. Since the internal
standards of the glyceride derivatives were labeled on the glycerol
moiety itself, the resulting product ions had the same m/z as the
unlabeled precursors. We investigated whether cross talk arising
from monitoring the same product ion m/z for both the analyte
and internal standard would affect instrument response. A
comparison of absolute peak areas of unlabeled 2-AG and its
internal standard 2-AGd5 showed that there was no significant
difference in area for either analyte when run separately, in a 1:1
ratio or in a 5:1 2-AG/2-AGd5 ratio. We therefore concluded that
cross talk was not a factor in the analysis.
Determining the optimal product ion for the SRM transitions
of the free fatty acids was difficult because of the absence of
chemical moieties whose cleavage would result in a predominant
fragment. In negative mode APCI, the product ion fragment we
monitored was the decarboxylation product of the fatty acid,
corresponding to a loss of 44 amu. The transitions worked well
for AA, DHA, and EPA, due to the relatively large number of
double bonds (4, 6, and 5, respectively) that could support a
negative charge. However, the sensitivity of PA, OA, and EA (0,
1, and 1 double bond, respectively) under these conditions was
too poor for this analysis. We therefore decided not to include
them in this study.
Although the free acids of palmitic, oleic, and eicosenoic acid
were not included in the analysis, their respective ethanolamide
5588 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Table 4. Endocannabinoid Stability in Ethanol and BSA (%) over 2-Months Storage at -80 °C and Four Freeze-Thaw
Cycles
ethanol single thaw
ethanol multiple thaw
week month month
BSA single thaw
BSA multiple thaw
day
1
week month month
day
1
day
1
week month month
day
1
week month month
analyte
1
1
2
1
1
2
1
1
2
1
1
2
AA
DHA
EPA
94.1
93.1
101.
95.3
94.5
101.3
96.8
99.7
88.3
57.7 109.6
96.0 102.7
55.4 100.4
118.0 95.2
93.9
93.9
56.8 105.0
91.4
94.3
96.6
99.3
93.5
95.2
52.5
109.0
99.4
107.1
93.9
94.9
94.3
99.6
90.7
90.6 101.4
104.2
96.4
93.3
98.5
98.7 103.5
96.2 100.5 101.5
81.3
AEA
100.7
101.4
97.8
98.8
98.4
99.4
96.2
98.6
98.9
97.3
98.7
90.3 106.4
88.8 97.8
92.6 102.5
102.4 95.6
99.1 101.1
95.9 102.7 100.6 103.1
98.6 103.1
94.8 102.8
89.9 100.4
87.4 103.2
97.7 103.5
96.5 102.4 100.6 100.3
91.6
88.3
88.5
91.3
97.3
102.1
PEA
97.6 100.9
96.7
94.9
95.0
90.2
97.9
97.5
99.6
99.3
99.6
98.8
86.2 100.1
90.8 100.6
98.9 106.2
96.1
98.8
92.6
90.0
98.8
97.9
95.6
97.7
99.6
97.1
OEA
DHEA
EPEA
EEA
99.0
98.7
96.1
91.5 102.2
89.9 99.1
95.2
97.3
92.0
99.1
107.0
88.1 104.2
94.2 97.1
110.2
103.7 102.1
105.0
89.4
97.4 100.5
2AG
2PG
2OG
93.9
97.6
98.5 101.9
99.2 96.8
95.1 105.3
107.8
89.3
93.9
98.4
98.4
97.6
98.5
83.4
92.4
96.0
99.4
109.7
95.3
84.6
98.5
89.0
98.3
98.9
98.9
87.5
91.9
102.0 92.9
93.2
99.2
98.7
84.
99.5
96.5
98.9
96.8
93.1
106.0
88.7
89.9
88.2
96.5 101.2
102.9 123.8
95.2 103.2 100.2
120.7
102.1 109.1 105.6
97.8 112.2 111.2
111.1 118.8
89.1
2DHG 101.9
94.0
88.2
95.4
96.4
97.0 104.2 105.6
89.6 102.9
89.3 98.6
98.2 114.6 100.3
2EPG
2EG
102.7 102.7
99.4 99.3
92.3
97.8
103.4
96.2
93.3
97.8
97.4 102.5
96.2 101.1
88.7 100.6
94.3 99.4
90.9
99.1
and glyceride derivatives were. The structure of each metabolite
and its internal standard are shown in Figure 1. In addition, the
text accompanying each structure includes its formula and
molecular weight as well as the SRM transition and collision
energy used in the analysis.
absolute peak areas for extracted/nonextracted samples were also
within 10-15% (not shown).
In order to confirm the validity of using BSA as a substitute
matrix for actual brain tissue during sample preparation, the
method of standard additions, which can quantify endogenous
analytes by direct internal calibration using the true sample matrix
rather than through external calibration, was performed using
AEA, 2-AG, and EPA. These analytes were chosen to represent
the entire group of endocannabinoid metabolites based on their
differences in relative abundance and chemical functionality.
Analyte peak areas at each level were plotted against the spike
amount and the resulting best fit curve was back-extrapolated to
the x-intercept to determine the endogenous level of analyte
present. The endogenous level was subtracted from the measured
spiked amount and then compared to the actual spiked amount
to determine the extraction efficiency. Table 2 reports the percent
recovery and the R2 value of the best-fit curve obtained for AEA,
2-AG, and EPA. The percent extraction recovery is an average of
triplicate measurements. The percent recoveries from rat brain
homogenate are slightly higher for the three analytes compared
to their recoveries from the BSA matrix. Direct comparison shows
that the recoveries are within 15% variation of each other and
demonstrates that it is reasonable to use BSA as an alternative
sample matrix to brain homogenate.
Validation. The endocannabinoids and related lipid mediators
are produced endogenously and are present in the brain and
plasma. Therefore, it is not possible to use plasma as the matrix
for preparing calibration standards, since the endocannabinoids’
baseline levels in plasma would alter the known calibration
amounts. To mimic as closely as possible the plasma and brain
matrixes, we chose to use a solution of BSA, the most abundant
protein in plasma, which had been stripped of its fatty acid
component. Based on results obtained from a Bradford assay
(Pierce Biotechnology, Rockford, IL), we found the level of total
protein in mouse brain homogenate to be 20 mg/mL. This level
was chosen as the concentration of “fatty acid-free” BSA to use
as our standard matrix. In order to ensure that there was no
contribution from the BSA matrix to the measured levels of
endocannabinoids, a series of blanks were run to ascertain possible
interference. The results for the ethanol blank, the BSA extracted
matrix blank, and a BSA extracted matrix spiked with internal
standard only (standard A) confirmed that the matrixes did not
contain any detectable levels of endocannabinoids.
The recovery of each analyte from the matrix following the
protein precipitation and liquid extraction steps was determined
in order to measure the efficiency of the extraction. Reference
samples were prepared with 100 pg/ul each of standard and
internal standard in ethanol and 20 mg/mL BSA matrix. The
ethanol reference standard was injected without extraction while
the BSA reference standard was extracted following the procedure
outlined in the Experimental Section. The amounts of nonex-
tracted and extracted reference standards were back-calculated
from the calibration curves for each analyte, and the percent
recovery was calculated from the extracted/nonextracted ratio.
Table 1 lists the percent recovery after sample workup for each
endocannabinoid studied. The recovery values ranged from 90.5
to 102.7%, indicating high extraction efficiency. Comparisons of
The linearity of the method over the concentration range of
5-500 pg/µL was determined by the calibration curves con-
structed for each analyte. Curves plotted the ratio of the analyte
area to its internal standard area against concentration. No
weighting factors were used to bias either end of the best-fit linear
curve through the data points. Regression analysis determined
R2 values of 0.99 or greater for each analyte. The lower limit of
quantitation, the amount of sample required to give a signal-to-
noise (S/N) ratio of 10 or greater, was determined to be 25 pg on
column for each analyte. Similarly, the limit of detection, the
amount of sample required to give a S/N ratio of 3 or better, was
∼10 pg on column for each analyte. Figure 2 shows the extracted
ion chromatogram for each transition monitored. The analytes
are displayed in the order of their elution from the column.
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5589

Figure 3. Extracted endocannabinoid MRM chromatograms from rat frontal cortex. Each chromatogram has been normalized to display the
peak of maximum intensity within the time window.
Arachidonic, docosahexaenoic, and eicosapentaenoic acids elute
within the first 6 min. The polarity switches from APCI- to APCI+
at 6 min to detect the ethanolamide and glyceride derivatives of
the six core fatty acids described in the Experimental Section.
Table 3 lists the accuracy and precision data acquired for
intraday and interday analyses. The accuracy and precision were
determined using QC samples prepared at four concentrations
spanning the entire concentration range measured by the method.
The concentrations of QC samples 1, 2, 3, and 4 were 7, 21, 120,
and 400 pg/ul, respectively. Triplicate injections of each QC
sample were made three times daily over the course of 24 h to
determine the method’s intraday variability. The QC sample mean
value for each run was obtained by averaging the back-calculated
amounts of the triplicate injections for each of the three analyses
made during 1 day. Accuracy was calculated as the percent
difference between the daily mean QC value and the true value.
Precision was calculated as the percent standard deviation of the
three daily values. To determine the interday variability of the
method, the accuracy and precision of runs 1-3 were averaged.
New calibration curves were run each day in conjunction with
the QC samples. The mid- to high-level QC samples (QC 2-QC
4) for each component had greater than 85% accuracy and
precision while the low QC 1 sample’s accuracy and precision was
greater than 80%. These levels meet the acceptance criteria
5590 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

Figure 4. Endocannabinoid metabolomic profile of rat frontal cortex.
outlined in the validation guidelines. The majority of samples had
accuracy and precision greater than 90%, indicating that the
method can be considered accurate and precise across the range
of concentrations used.
nolamides into carboxylic acids and ethanolamine.27 FAAH acts
as one of the main regulatory enzymes associated with the
hydrolysis of the endocannabinoid anandamide.28 Treatment with
AM374 was expected to inhibit the degradation of anandamide
within the brain and elevate the level of anandamide, compared
to vehicle treatment. Figure 3 shows an example of the extracted
MRM chromatograms obtained from rat frontal cortex homoge-
nate. Matching each endocannabinoid’s retention time to its
internal standard was performed to ensure that the correct peak
in each chromatogram was integrated properly. Small deviations
in retention times between reference standards (Figure 2) and
brain homogenate (Figure 3) are paralleled by corresponding
deviations of each endocannabinoid’s internal standard. The acyl
glycerol derivatives exhibit a slightly distorted peak shape due to
the partial chromatographic resolution of the sn-2 isomer from
the sn-1 or sn-3 isomers. Since this method does not fully resolve
each isomer, reported acyl glycerol levels include quantification
of all isomers. Figure 4 shows the profile of the endocannabinoid
metabolome as measured in the frontal cortex of the rat vehicle
samples. The levels of individual metabolites vary over 4 orders
of magnitude and as such are presented on a logarithmic scale.
AEA was the lowest quantifiable metabolite at 1.7 ng/g of wet
tissue, while the level of AA was 18.5 µg/g. EEA was detected
but was below the limit of quantitation for the analysis. EPEA was
not detected at all. The levels of AEA, 2-AG, OEA, and PEA found
in mouse brain are within the range reported by others.19,29-31
The stability of the endocannabinoid metabolites was studied
in both ethanol and BSA over 2 months at -80 °C. Aliquots from
each matrix were analyzed following 1 day, one week, 1 month,
and 2 months. For each time point, two sets of samples initially
were prepared that served as single and multiple thaw sets.
Therefore, the analysis on day one consisted of two sets each
thawed just once. Within 2 months, the multiple thaw sample set
had cycled through four freeze/thaws, corresponding to one
freeze/thaw for each time point analyzed. The single thaw set
had been maintained at -80 °C for the full 2 months. The results
are listed in Table 4, which shows the percent of each metabolite
remaining in solution for the specific matrix, time, and freeze/
thaw point. For each time point, fresh calibration curves and
reference samples for each metabolite were compared to the
stability samples. The stability samples were spiked with fresh
internal standard prior to analysis for an accurate reflection of
peak area ratio differences. The majority of the metabolites at
each time point were within 15% of the reference sample level.
The one metabolite that degraded significantly by the second
month was arachidonic acid. In both ethanol and BSA, the levels
had dropped to between 50 and 60% of the reference sample. The
levels were similar in the multiple thaw samples. Due to the
instability of AA at 2 months, the endocannabinoid stock solu-
tions were prepared fresh once per month to maintain sample
integrity. The metabolites are stable through four cycles of
freeze-thaw.
(27) Deutsch, D. G.; Lin, S.; Hill, W. A.; Morse, K. L.; Salehani, D.; Arreaza, G.;
Omeir, R. L.; Makriyannis, A. Biochem. Biophys. Res. Commun. 1997, 231
(1), 217-21.
(28) Egertova, M.; Cravatt, B. F.; Elphick, M. R. Neuroscience 2003, 119 (2),
481-96.
(29) Baker, D.; Pryce, G.; Croxford, J. L.; Brown, P.; Pertwee, R. G.; Makriyannis,
A.; Khanolkar, A.; Layward, L.; Fezza, F.; Bisogno, T.; Di Marzo, V. FASEB
J. 2001, 15 (2), 300-2.
(30) Cravatt, B. F.; Demarest, K.; Patricelli, M. P.; Bracey, M. H.; Giang, D. K.;
Martin, B. R.; Lichtman, A. H. Proc. Natl. Acad. Sci. U.S.A. 2001, 98 (16),
9371-6.
Analysis of Biological Samples. To demonstrate the meth-
od’s applicability for quantifying the endocannabinoid metabolome,
rat brain frontal cortex regions were analyzed. Briefly, two groups
of Sprague-Dawley rats (n ) 3 in each group) were dosed
intraperitoneally with either vehicle or 8 mg/kg palmitylsulfonyl
fluoride (AM374). AM374 is a known inhibitor of the enzyme
FAAH, fatty acid amide hydrolase, that hydrolyzes N-acyl etha-
(31) Franklin, A.; Parmentier-Batteur, S.; Walter, L.; Greenberg, D. A.; Stella, N.
J. Neurosci. 2003, 23 (21), 7767-75.
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5591

Figure 5. Extracted chromatogram of AEA in rat frontal cortex treated with (A) vehicle or with (B) FAAH inhibitor AM374.
The levels of the other acyl glycerol and acyl ethanolamine
metabolites determined in this study have not been reported. This
profile shows that metabolite levels for an individual core structure
are lowest for the ethanolamide derivative and increase for the
glyceride derivatives and free acids. Due to the greater abundance
of the N-acyl glycerols, such as 2-AG or 2-PG, over the N-acyl
ethanolamines, these glyceride derivatives may contribute more
to the physiological effects of endocannabinoids than the N-acyl
ethanolamines (such as AEA) even though cannabinoid receptor
binding affinity may favor the lesser component.
Figure 5A and Figure 5B are representative extracted ion
chromatograms of AEA from the rat frontal cortex analysis that
correspond to treatments with the vehicle and AM374, respec-
tively. The sample time point in Figure 5B is 3 h after administer-
ing AM374. A comparison of the peak areas shows a 9-fold
increase in AEA levels between the two samples. The calculated
AEA levels between the untreated and treated frontal cortex
samples were 0.64 and 7.50ng/g, respectively. These data indicate
that treating with the FAAH inhibitor AM374 causes a significant
increase in AEA levels in the frontal cortex. A comprehensive
study is currently being completed that monitors endocannabinoid
levels following AM373 treatment at several times and at several
locations in rat brain using the presently developed method.
Correlations between the neuroprotective properties of AM374,
animal behavior, and changes in the entire endocannabinoid
metabolome will be presented.
The paper reports on the simultaneous analysis of eight metabo-
lites using LC-ESI-MS but employs only two internal standards
for quantifying the entire set of metabolites. Differences in the
selected metabolites’ ionization efficiency due to differences in
chemical structure and chromatographic elution favor the use of
separate internal standards for each metabolite for improved
quantitative accuracy. The current endeavor effectively doubles
the number of endocannabinoids that can be quantified simulta-
neously, allowing for a more comprehensive endocannabinoid
metabolomic profile to be established. The validation criteria
established for bioanalytical methodology have been met for
several important parameters. Linearity based upon regression
constants greater than 0.99 were obtained for each metabolite,
over the calibration range of 25 pg to 2.5 ng on column. The
absence of interferences tested in a series of blanks at the chosen
MRM transitions demonstrated the specificity of the methodology.
The accuracy, precision, and ruggedness of the method were
established by examining the reproducibility of several QC
samples over extended times and conditions. Finally an endocan-
nabinoid profile for the rat frontal cortex has been established
and the variations in anandamide level due to treatment with the
FAAH inhibitor AM374 have been quantified successfully. The
potential benefit of establishing known endocannabinoid profiles
extends far beyond the rat model used in this study. The
endocannabinoid system may play a significant role in human
disease pathologies ranging from neurological disorders such as
schizophrenia and postpartum depression to cannabinoid abuse
and dependence. Additionally, several cannabinergic xenobiotics
have been developed by this group and others that target key
endocannabinoid proteins and modulate the composition of the
endocannabinoid metabolome. Thus, determination of the levels
of endocannabinoid metabolites under diverse physiological condi-
tions provides an excellent window of information on the physi-
ological system being studied. As novel endocannabinoids con-
CONCLUSIONS
In this paper, we describe the development, validation, and
application of a quantitative assay to profile the levels of several
members of the endocannabinoid metabolome by LC-APCI-MS-
MS. The endocannabinoids are an important group of endogenous
ligands for the cannabinoid system of receptors and enzymes
whose complete biological significance is currently under intense
investigation. The need for a thorough evaluation of the endocan-
nabinoid metabolome was also illustrated in a recent publication,
which came to our attention during the preparation of this article.32
(32) Richardson, D.; Ortori, C. A.; Chapman, V.; Kendall, D. A.; Barrett, D. A.
Anal. Biochem. 2007, 360 (2), 216-26.
5592 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

tinue to be discovered, it is necessary to further develop the
quantitation methodology. We are currently working on high-
throughput methodologies with greater sensitivity to tackle the
rapidly expanding endocannabinoid metabolome.
SUPPORTING INFORMATION AVAILABLE
NMR structural assignments for all endocannabinoid metabo-
lites synthesized in-house. This material is available free of charge
via the Internet at http://pubs.acs.org.
Received for review December 20, 2006. Accepted May
24, 2007.
ACKNOWLEDGMENT
This research was supported by grants from the National
Institutes of Health, DA-91587215 and DA-91583801.
AC0624086
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5593