Multi-Omic Analysis Reveals Cannabidiol Disruption of Cholesterol Homeostasis
in Human Cell Lines
Steven E. Guard1,5, Douglas A. Chapnick2,5, Zachary Poss1, Christopher C. Ebmeier1,
Jeremy Jacobsen1, Travis Nemkov3, Kerri A. Ball1, Kristofor J. Webb1, Helen L.
Simpson1 Stephen Coleman1, Eric Bunker2, Adrian Ramirez2, Julie A. Reisz3, Robert
Sievers4, Michael H.B. Stowell1, Angelo D’Alessandro3, Xuedong Liu2, William M.
1 Dept. of Molecular, Cellular & Developmental Biology, University of Colorado Boulder
2 Dept. of Biochemistry, University of Colorado Boulder
3 Dept. of Biochemistry and Molecular Genetics, University of Colorado Denver
4 Dept. of Chemistry and Cooperative Institute for Research in Environmental Sciences,
University of Colorado Boulder
5 These authors contributed equally
6 Lead Contact
Cannabidiol (CBD) is FDA-approved for treatment of drug-intractable forms of pediatric
epilepsy, yet the mechanisms that underlie its efficacy remain unclear. Myriad protein
targets of CBD have been reported, suggesting a pleiotropic pharmacology. Here, we
report a systems-level analysis of CBD action in human cell lines using temporally-
resolved multi-omic profiling and biosensor screening. CDB treatment resulted in a
chronic rise in cytosolic calcium and activated AMPK signaling within two hours.
Subcellular profiling of proteins, metabolites, and mRNA transcripts identified CBD-
dependent activation of cholesterol biosynthesis, transport and storage. We found that
CBD incorporates into cellular membranes, alters cholesterol chemical activity, and
increases lipid order. CBD-induced apoptosis in multiple human cell lines was rescued
by inhibition of cholesterol synthesis, and potentiated by compounds that disrupt
cholesterol trafficking and storage. Our data point to pharmacological interaction of CBD
with cholesterol homeostasis pathways, with potential implications in its therapeutic use.
Keywords: cannabidiol, CBD, systems pharmacology, drug discovery, cholesterol,
proteomics, transcriptomics, lipidomics, metabolomics, phosphoproteomics, FRET
biosensor, high-content screening, multi-omics
The non-psychoactive cannabinoid, cannabidiol (CBD), was recently approved by the
FDA for treatment of two drug-resistant epileptic disorders, Lennox-Gastaut syndrome
and Dravet syndrome (Devinsky et al., 2017; Miller et al., 2020; Thiele et al., 2018).
CBD has an encouraging safety profile across multiple human clinical trials, which has
.CC-BY 4.0 International licensewas not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (whichthis version posted June 4, 2020. . https://doi.org/10.1101/2020.06.03.130864doi: bioRxiv preprint
fueled increased interest in CBD as a therapeutic in other disorders (Iffland and
Grotenhermen, 2017). CBD has shown promise as a potential therapeutic from animal
model studies of a diverse array of conditions, including oxidative stress and
inflammation, chemotherapy associated nephrotoxicity (Pan et al., 2009), colitis (De
Filippis et al., 2011), cancer (Hinz and Ramer, 2019; Kenyon et al., 2018; Massi et al.,
2013; McAllister et al., 2015), neuroinflammation (Esposito et al., 2011),
cardiomyopathy, and diabetic complications (Rajesh et al., 2010). However, CBD
treatment has shown an increased risk for liver injury, particularly for patients taking
valproic acid (Thiele et al., 2018). Remarkably, despite its promise as a broad
therapeutic, the molecular mechanisms that underlie the efficacy and the potential
toxicities of CBD in humans remain poorly understood.
CBD has weak affinity (> 3-10 μM) for the endocannabinoid receptor CB1, and is thus
devoid of the psychotropic effects associated with its better known structural isomer,
tetrahydrocannabinol (THC) (McPartland et al., 2015). More than 65 protein targets of
CBD have been proposed, 22 of which are membrane-localized channels and receptors
(Ahrens et al., 2009; Ibeas Bih et al., 2015; Lauckner et al., 2008; Whyte et al., 2009).
Notably, CBD inhibits voltage-dependent sodium currents mediated by the NaV1.1
sodium channel (Ghovanloo et al., 2018), mutations in which elicit epilepsy syndromes,
including Dravet syndrome (Dravet, 2011). CBD was also shown to inhibit voltage-
dependent ion currents of six other human sodium channels, the Kv2.1 potassium
channel, and even a bacterial sodium channel, with IC50 values 1-3 μM (Ghovanloo et
al., 2018). Many proposed targets are calcium channels or receptors that regulate
calcium, including T-type calcium channels (Ross et al., 2008), voltage-dependent anion
channel 1 (VDAC1) (Rimmerman et al., 2013), G protein-coupled receptor 55 (GPR55)
(Lauckner et al., 2008), a voltage-gated calcium channel Cav3.x (Ross et al., 2008), and
transient receptor potential cation channels 1-4 (TRPV1-4) (Ibeas Bih et al., 2015).
Postsynaptic calcium mobilization has been proposed as a mechanism to explain the
anticonvulsant activity of CBD (Gray and Whalley, 2020). The ability of CBD to
modulate many structurally diverse membrane channels and receptors has motivated
suggestions that CBD, in a similar manner to other amphiphilic molecules such as
capsaicin, may act indirectly on membrane proteins through alteration of the biophysical
properties of the lipid bilayer (Ghovanloo et al., 2018; Lundbaek et al., 2005; Lundbæk
et al., 1996; Watkins, 2019).
Comparatively less is known about intracellular targets and pathways that potentially
mediate CBD action in human cells. In microglial cells, CBD has anti-inflammatory
activity, and upregulates mRNA transcripts involved in fatty acid metabolism and
cholesterol biosynthesis (Rimmerman et al., 2011). In adipocytes, CBD leads to
accumulation of triglyceride species, concomitant with phosphorylation changes of
upstream lipid metabolism enzymes, and functionally unrelated proteins including
CREB, AMPKA2 and HSP60 (Silvestri et al., 2015). In mice, CBD attenuates the liver
steatosis and metabolic dysregulation associated with chronic alcohol feeding, further
implicating lipid-associated metabolic changes induced by CBD (Wang et al., 2017).
The perturbation of transcripts, proteins and metabolites in these studies demonstrates
the pleiotropic effects of CBD on diverse molecular classes, and highlights the need for
systems-based approaches for examining its mode of action.
.CC-BY 4.0 International licensewas not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (whichthis version posted June 4, 2020. . https://doi.org/10.1101/2020.06.03.130864doi: bioRxiv preprint
Recent advances in mass spectrometry-based omics has enabled nearly
comprehensive identification and quantification of cellular proteomes and metabolomes
(Blum et al., 2018). Multi-omic profiling strategies that combine mass spectrometry
based proteomics with next generation sequencing-based transcriptomics can reveal
critical and unexpected insights into the mechanisms of drug action in human cell lines
(Hafner et al., 2019; Norris et al., 2017). In this study, we used multi-omic profiling, in
combination with FRET-based biosensor microscopy screening, to examine the mode of
CBD action in human neuroblastoma and keratinocyte cells. CDB treatment activated
5'-AMP-activated protein kinase (AMPK) and cytosolic calcium signaling within three
hours in multiple cell types. In SK-N-BE(2) neuroblastoma cells, proteomics,
transcriptomics and lipidomics revealed a concerted upregulation of transcripts and
proteins involved in cholesterol import and biosynthesis. We show that CBD sensitized
human cells to drugs that interfere with cholesterol trafficking and storage, including an
inhibitor of Niemann-Pick C1 (NPC1), U18666A. Using model membranes, we found
that CBD increased the chemical activity of cholesterol as a substrate for cholesterol
oxidase, while also reducing the lateral diffusion of cholesterol. Our data reveal that a
dominant action of CBD involves partitioning into cellular membranes, which leads to
the disruption of cholesterol homeostasis. This observation likely underlies the
downstream effects of increased apoptosis and the widely reported pharmacological
pleiotropy of CBD.
FRET-based sensor array reveals CBD response dynamics
To identify time-ordered molecular events initiated by CBD, we performed temporal
multi-omic profiling of CBD treated human neuroblastoma cells, using proteomics,
phosphoproteomics, RNA-sequencing (RNAseq) and metabolomics. The dynamics of
metabolite, RNA, and protein changes in response to drug perturbation could span time
scales ranging from seconds to days, presenting a challenge for selecting appropriate
time points in multi-omic analysis. To identify the optimal time points and CBD dose, we
monitored a panel of Förster resonance energy transfer (FRET) sensors over time with
high-content imaging in human SK-N-BE(2) neuroblastoma cells and HaCaT human
keratinocyte cells. Transgenic lines were generated, each expressing a genetically
encoded FRET biosensor gene capable of reporting the activity of a cellular activity
(Chapnick et al., 2019). Sensors were selected to profile a broad range of molecular
activities measuring abundances changes in metabolites, secondary messengers as
well as kinase and protease activities (Table S1A).
FRET ratios were measured in a time course following either vehicle or CBD across a
range of doses from 0 to 100 µM (Figure S1A). At each time point, we fit a log-logistic
function with the FRET sensor data to estimate EC50 values and quantify the dose-
dependency for each sensor over time. We found that cytosolic calcium, plasma
membrane charge, AMPK activity, ERK activity, and glucose abundance exhibited the
most significant dose-dependent changes. A number of sensors showed time
dependent responses at various doses of CBD, but showed poor dose-dependency (R2
<= 0.75), and thus were not used for dose and time range analysis. The EC50
distribution of CBD dose across all time points and biosensors displayed a median of
.CC-BY 4.0 International licensewas not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (whichthis version posted June 4, 2020. . https://doi.org/10.1101/2020.06.03.130864doi: bioRxiv preprint
8.5 µM for SK-N-BE(2) cells. We found, however, that 20 µM CBD was required for
activation of FRET sensors for which a EC50 could be calculated at early time points,
including cytosolic calcium, AMPK, and plasma membrane charge (Figure 1A). SK-N-
Figure 1. FRET biosensor screening strategy for dose and time selection of multi-omic CBD
perturbation analysis. (A) EC50 distribution of CBD treatment across all sensors and timepoints (See
also Table S1 for list of sensors). Dose response curves were fit to determine EC50 values (R2 values >
0.75) (B) Heat maps of FRET biosensor responses to CBD treatment for cytosolic Ca2+ and AMPK
activity in SK-N-BE(2) cells, displaying FRET ratio over time at CBD doses from 343 nM to 100 µM. (C)
Relative abundance of CBD over time from metabolomic profiling of SK-N-BE(2) cells treated with 20
µM CBD. (D,E) Time course schematic of multi-omic experimental strategy.
BE(2) cells were less sensitive to CBD treatment compared with HaCaT cells, and
displayed a higher degree of dose-dependency in FRET sensor activation over time
(Figure S1A). We therefore selected a CBD dose of 20 µM, and SK-N-BE(2) cells for
subsequent multi-omic experiments.
After treatment with 20 µM CBD, the earliest events detected by the biosensors showing
dose-dependence were an increase in cytosolic calcium at 3 hours followed closely by
AMPK activation (Figure 1B). AMPK can be allosterically activated by AMP when
AMP:ATP ratios increase, or by the upstream kinases, Ca2+/Calmodulin-dependent
protein kinase kinase β (CaMKKβ) and LKB1 (Hawley et al., 2003, 2005; Shaw et al.,
2004). CaMKKβ increases the activity of AMPK through direct interactions with its
kinase domain, driving downstream secondary calcium signaling events (Hawley et al.,
2005; Woods et al., 2005). This calcium FRET sensor response is consistent with
previous reports of CBD treatment driving an increase in cytosolic calcium through
either TRPM8, TRPV receptors, or Voltage-dependent T type receptors (Ibeas Bih et
al., 2015; Rimmerman et al., 2013).
Following dose and cell line selection, we monitored the kinetics of cellular uptake of
CBD in SK-N-BE(2) cells. The relative abundance of intracellular CBD was quantified by
mass spectrometry in a time course from 30 seconds to 72 hours. CBD was detected in
cells as early as 30 seconds but did not reach steady state until 80 minutes post
treatment (Figure 1C). Based on these kinetics of CBD uptake, we performed a set of
multi-omic experiments to examine the temporal response of SK-N-BE(2) cells to CBD
treatment, from minutes to days, using global metabolomics, lipidomics,
phosphoproteomics, subcellular proteomics, and transcriptomics (Figure 1D), resulting
in the detection of > 42,000 phosphorylated peptides, 8,359 proteins, 21,517 gene
transcripts and 16,129 metabolic features (Figure 1E).
CBD activates AMPK signaling and downstream substrate phosphorylation
We performed SILAC phosphoproteomics to quantify changes in phosphorylation in
response to CBD treatment at 10 minutes, 1-hour and 3-hour time points. At the 10-
minute time point, only five significantly changing phosphorylation sites were observed
(q < 0.05 and |log2 ratio| > 0.5) (Figure S2A). However, the number of significantly
changing sites increased to 154 by 1 hour (Figure 2A), mirroring the kinetics of CBD
uptake into cells between 40 and 80 minutes (Figure 1C). At both 1 hour and 3-hour
timepoints, significantly changing phosphorylation sites were enriched in AMPK
signaling proteins. (Figure 2A, 2B and S2B). The canonical phosphorylation motif of
high confidence AMPK substrates has been identified as L-X-R-X-X-(pS/pT)-X-X-X-L
based on peptide amino acid frequencies (Dale et al., 1995; Gwinn et al., 2008;
Schaffer et al., 2015). We found that AMPK motifs were significantly enriched in CBD-
responsive phosphorylation site sequences at 1- and 3-hours using sequence motif
analysis, including L-X-R-X-X-pS and R-X-X-pS-X-X-X-L (Figure S2C and S2D). An
average of these motifs emerges in the sequence logo at these time points (Figure 2C
Figure 2. CBD increases phosphorylation of AMPK signaling proteins at early time points. (A-
B) Volcano plots of quantified phosphorylation sites at 1 and 3 hours post treatment with 20 µM CBD,
showing significance of differential change on y-axis and log2(CBD/DMSO) ratio on x-axis. Red:
adjusted p < 0.05, |log2(CBD/DMSO) | > 0.5. Sites on AMPK proteins involved in AMPK signaling with
adjusted p-value < 0.05 are annotated in white boxes. (C- D) Phospho motif enrichment from
phosphorylation sites identified as significantly changing in CBD treated cells vs vehicle control at 1-
and 3-hours post CBD treatment. The AMPK averaged motif is displayed beneath the sequence logo
(Schaffer et al., 2015) (E) Overlay of significantly changing phosphorylated sites onto an AMPK
signaling diagram. Asterisks signify phosphorylation sites with a known upstream kinase (F-G)
Seahorse extracellular flux measurement of oxygen consumption rate (OCR) and extracellular
acidification rate (ECAR), * p < 0.05; ** p < 0.01. Stressed condition: Oligomycin treatment (1 µM)/
FCCP treatment (1 µM).
Of the observed phosphorylation sites on proteins involved in AMPK signaling, several
of the CBD-responsive events are annotated with biological function. We observed
increased phosphorylation of S108 within the beta subunit of AMPK at the 1-hour and 3-
hour timepoints. Phosphorylation at this site drives a conformational change in the
AMPK complex resulting in stabilization of active kinase by preventing
dephosphorylation of the activation site at T172 (Li et al., 2015). In agreement with an
increase in AMPK activity after 1 hour of CBD treatment, we found increased
phosphorylation of S80 on acetyl-CoA Carboxylase (ACACA), a known AMPK
phosphorylation site (Figure 2E) (Carlson and Kim, 1973; Munday, 2002). ACACA
catalyzes the rate-limiting step of fatty acid synthesis, and is deactivated by AMPK
phosphorylation of S80. Phosphorylation of ACACA on S80 results in reduced
conversion of acetyl-CoA into malonyl-CoA, reducing carbon flux through fatty acid
synthesis, and increasing catabolic fatty acid β-oxidation (Fediuc et al., 2006;
McFadden and Corl, 2009). In line with these findings, we observed decreased flux of
carbon into de novo synthesized fatty acids (Figure S2E). We found significantly
decreased levels of short and medium chain, but not long chain acylcarnitines in the
CBD-treated cells, indicating that fatty acid mobilization is comparable in the two
groups, but more rapidly fluxed through fatty acid β-oxidation upon treatment with CBD
We also identified increased phosphorylation of the translation elongation factor, EEF2,
on T56 with CBD treatment at 1 and 3 hours. EEF2 T56 phosphorylation is sufficient to
inhibit the GTP-dependent ribosomal translocation step during translational elongation,
which suggests upstream activity of AMPK and EEF2K (Ryazanov et al., 1988).
Together, these observations predict downstream alterations of both protein and fatty
acid synthesis rates downstream of AMPK signaling.
Agreement between phosphoproteomics and FRET sensor activity indicated that AMPK
is activated by CBD treatment, but did not reveal the mechanism of activation.
Upstream activation of AMPK can be initiated through conformational changes driven by
cellular AMP:ATP ratio, or through calcium dependent phosphorylation events (Hardie
et al., 1999; Woods et al., 2005). To test whether CBD treatment acutely alters cellular
energy status, we measured the oxygen consumption rate and extracellular acidification
rate of CBD treated cells using a Seahorse extracellular flux assay. Treatment of SK-N-
BE(2) cells with 20 µM CBD led to decreased levels of basal oxygen consumption by 24
hours, with little change at 2 hours post-drug treatment (Figure 2F). Basal extracellular
acidification rate remained unchanged (Figure 2G). Consistently, we observed
comparable rates of lactate production, but decreased carbon flux into TCA cycle
metabolites in cells treated with CBD (Figure S2G). These results suggest that CBD-
treated cells have decreased ATP production by mitochondrial respiration with little to
no compensation by glycolysis, which may sustain AMPK activation at late time points.
While we do not have direct evidence of the mechanism by which AMPK is activated
between 1-3 hours, the most likely explanation is that in the absence of compromised
ATP production, calcium influx into the cytoplasm leads to activation of AMPK by
calcium activated upstream kinases such as CAMKKβ.
Figure 3. CBD treatment upregulates cholesterol biosynthesis enzymes and translocation of
metabolic proteins. (A) Compositionally distinct subcellular proteomic fractions were fractionated by
differential centrifugation and pH. The “cytosolic” fraction is enriched in soluble protein, the “nuclear”
fraction is enriched in insoluble subnuclear compartments: condensed chromosome, spindles,
spliceosomal complex etc., and the “membrane” fraction is enriched in membrane and mitochondrial
related proteins. (See also Figure S3C -S3E) (B) Frequency of significantly changing proteomic events
over time. (C) Anticorrelated proteins between proteomic fractions over time. PCA dimensionality
reduction was used to decrease the impact of noisy signal contribution. Correlation between fractions, r
< -0.8 was required. A large proportion of proteins listed above are known to compartmentalize in the
mitochondria indicating protein shuttling or mitochondrial detachment/ attachment. (See also Figure
S3F) (D) Proteins and mRNA transcripts that change significantly with CBD and map to the indicated
gene ontology annotations that showed significant enrichment of differential proteins (Methods).
CBD upregulates transcripts and translocation of proteins involved in cholesterol
To identify time-dependent changes in subcellular localization of proteins, we developed
a pH-dependent cell fractionation scheme using differential centrifugation (Figure 3A;
Star Methods). The resulting ‘cytosolic’ fraction is enriched in soluble proteins from the
cytosol, nucleus as well as various luminal compartments through the cell
(mitochondrial, vesical, etc.) (Figure 3SA). The first insoluble fraction, labeled as
‘membrane’, contains proteins from both the mitochondrial and plasma membrane,
while the second insoluble fraction is highly enriched in insoluble nuclear components
such as condensed chromatin, spindles and nuclear speckles (Figure S3B and S3C).
Principal component analysis (PCA) of these fractions revealed three compositionally
distinct portions of the proteome, with each of these fractions exhibiting a time-
dependent separation in response to CBD treatment. (Figure S3D and S3E). However,
the membranous and nuclear fractions remain very similar in PCA space until 12 hour
and later time points, suggesting relatively slow kinetics of protein regulation in
response to CBD. Consistent with this observation, the frequency of significant events
across fractions are limited at time points prior to 12 hours but increase dramatically to
hundreds of proteins at timepoints between 15 to 72 hours. (Figure 3B).
Anticorrelation between the time-dependent responses for a protein in different
subcellular fractions would be expected for proteins translocating between cellular
compartments in response to CBD. To identify potential translocation events, we
calculated the Pearson’s correlation coefficient between the temporal profiles of each of
that protein’s subcellular fractions. We found 30 proteins with highly anti-correlated
subcellular profiles (Figure 3C). Notably, Hexokinase1 (HK1) is observed to decrease in
the membrane fraction and increase in the nuclear fraction (Figure 3C and S3F). HK1
detachment from the outer mitochondrial membrane results in severely decreased
activity to convert glucose to glucose 6-P (Berger et al., 1946). Consistently, CBD-
treated cells similarly demonstrated decreased levels of glucose-6-phosphate during
later timepoints of CBD treatment (Figure S3G). HK1 detachment decouples glycolysis
from mitochondrial respiration, and can alter the overall balance of energy metabolism
in the cell. This translocation event is consistent with decreased cellular respiration in
response to CBD treatment (Figure 2F) and previous reports of CBD-induced
mitochondrial dysfunction in neuroblastoma cells (Alharris et al., 2019).
To identify CBD-dependent changes in transcript abundance, we performed an RNA-
seq experiment, comparing SK-N-BE(2) cells treated with 20 µM CBD or vehicle for 3,
6, 12 and 24 hours. We identified 4,118 differentially expressed genes in CBD treated
cells that were significant in at least 1 time point with a false discovery rate (FDR) <1%.
204 of these genes displayed transcript abundances with a |log2 ratio| ≥1 (Figure 1E).
To identify potential transcription factor specific responses that explain mRNA transcript
changes, we performed upstream regulator analysis on significantly changing
transcripts (Krämer et al., 2014). The most enriched transcription factors for increasing
transcripts shared oxidative stress as a stimulus and included ATF4, NFE2L2, SP1
(Figure S3H) (Blais et al., 2004; Dasari et al., 2006; Venugopal and Jaiswal, 1998;
Wang and Semenza, 1993). Along similar time scales, CBD-treated cells showed an
accumulation of the principal cellular antioxidant glutathione, indicating an upregulation
in oxidative stress response (Figure S3G).
We merged differentially expressed transcript and protein identifications by gene name
and assigned gene ontology annotations using REVIGO pathway analysis (Figure 3D)
(Supek et al., 2011). CBD responsive events were enriched in translation, ER stress
response, metal ion response, and cholesterol biosynthesis (adj. p < 0.01). While many
of these annotations are supported by either the transcriptome or proteome,
dysregulation of cholesterol metabolism is supported by both. Within the cholesterol
biosynthetic pathway ontology enrichment, 17 proteins displayed significant abundance
changes that increased over time, including several key regulatory proteins. The rate
limiting enzyme in cholesterol synthesis, HMGCR, increased as much as 303%, with
increased transcript abundance by 6 hours. Negative regulators of this rate limiting
enzyme were observed to decrease, including a 40% decrease in SOD1, consistent
with decreased repression of HMGCR transcription (De Felice et al., 2004). The
enzyme catalyzing the conversion of desmosterol into cholesterol in the terminal step in
cholesterol biosynthesis, DHCR24, increased by 43% on the protein level in the
“membrane” fraction (Figure 3D). Together, the proteomic and transcriptomic data point
to a concerted response in cholesterol homeostasis pathways in response to CBD
treatment, and predict an upregulation of cholesterol biosynthesis capacity.
CBD treatment results in accumulation of cholesterol biosynthesis intermediates
and esterified cholesterol
Proteomic and transcriptomic analyses revealed a concerted CBD-induced upregulation
of cholesterol biosynthesis machinery. These findings raised the question of whether
CBD treatment leads to alterations in lipid and cholesterol metabolism (the latter
pathway depicted in Figure 4A). We used mass spectrometry-based lipidomics to
quantify the effect of CBD on lipids and sterols. Vehicle and CBD exposed cells were
pulse labeled with [U-13C6]-D-Glucose for 24 hours and harvested using methanol
extraction. Cholesterol biosynthetic flux was quantified by mass spectrometry analysis
of 13C incorporation into biosynthetic intermediates. We found that cholesterol
precursors accumulated in CBD exposed cells (Figure 4B and S4A), while labeled and
total cholesterol itself decreased modestly (Figure 4B and S4B). This effect of CBD on
total cellular cholesterol was confirmed using an Amplex Red cholesterol assay (Figure
Intracellular cholesterol is stored in lipid droplets after esterification with long-chain fatty
acids by ER-resident enzymes (Brown et al., 1980). We detected accumulation of
multiple species of cholesteryl esters with various chain lengths and acyl-chain
saturation (Figure 4C). Upregulation of cholesterol biosynthesis enzymes, together with
increased abundance of metabolic precursors, suggest that CBD leads to increased
production and storage of cholesterol.
Figure 4. Cholesterol biosynthesis precursors and cholesterol esters accumulate upon CBD
treatment. (A) Pathway diagram of cholesterol biosynthesis. Quantified metabolic intermediates
are outlined in black bold, enzymes showing significant increase over time with CBD in proteomic
analysis are shown bold red (Figure 3D). Dashed arrows indicate multiple intermediate steps that were
not identified in the pathway. (B) D-glucose (U-13C6) metabolically labeled cholesterol biosynthesis
precursors at 24 hours post 20 µM CBD treatment. (Student’s T Test : *p< 0.05, **p<0.01, ***p<0.001)
(See also Figure S4) (C-E) Total abundance of lipids quantified by LC/MS/MS from lipid extracts of SK-
N-BE(2) cells treated with vehicle or 20 µM CBD. Cholesteryl esters (CE), free head groups and
phospholipids identified by mass spectrometry are displayed. Phosphatidylcholine(PC);
Phosphatidylethanolamine (PE); lysophosphatidylethanolamine (LPE) (Student’s t-test: *p< 0.05,
Due to the requirement of acyl-CoA precursors in cholesterol esterification (Chang et
al., 2009), we surveyed our dataset for evidence of fatty acid utilization. We found that
CBD treatment led to the accumulation of metabolites derived from phospholipid head-
groups (Figure 4D and S4D), including ethanolamine phosphate, a product of
sphingosine catabolism via the enzyme S1P lyase (SGPL1) (Serra and Saba, 2010).
These data suggest that both free and SGPL1-generated fatty acids may be utilized for
cholesterol esterification. We also surveyed the cellular abundance of all detectable
species of phosphatidylcholine and phosphatidylethanolamine in cell extracts and found
a variety of phospholipids that display reduced abundance in CBD treated versus
vehicle treated cells (Figure 4E). These observations are consistent with previous
studies showing catabolic breakdown of phospholipids upstream of SGPL1 activity
(Aguilar and Saba, 2012). Thus, we found multiple lines of evidence consistent with
activation of cholesterol esterification in the CBD response.
CBD increases storage and transport of cholesterol
Cholesterol is a critical structural component of cellular membranes. Alterations in
cholesterol abundance can lead to severe cellular phenotypes that include
mitochondrial dysfunction and apoptosis (Zhao et al., 2010). Disruption in cholesterol
trafficking is a hallmark of Niemann-Pick Type C disease (Chang et al., 2005), and can
drive sodium channel-mediated inflammatory pain in animal models (Amsalem et al.,
2018). To explore the phenotypic implications of CBD disruption of cholesterol
homeostasis, we tested whether CBD-induced cell death required upregulated
cholesterol synthesis. Earlier experiments indicated that 40 µM CBD can induce cell
death after 24 hours in SK-N-BE(2) cells (Figure 5A). SK-N-BE(2) cells were exposed
to increasing concentrations of CBD in the presence or absence of the cholesterol
biosynthesis inhibitor, atorvastatin, and analyzed for apoptosis using CellEvent caspase
3/7 dyes and live-cell fluorescence imaging. At 15 hours, 100 µM CBD leads to
apoptosis of 50% of SK-N-BE(2) cells. Co-treatment of SK-N-BE(2) cells with CBD and
atorvastatin reduced apoptosis by approximately 2-fold (Figure 5B). This atorvastatin-
dependent rescue of CBD-induced apoptosis was far more pronounced in human
HaCaT keratinocytes (Figure S5A), which are highly sensitive to cholesterol
perturbation (Bang et al., 2005). Further, CBD exposed SK-N-BE(2) and HaCaT cells
show an increase in apoptosis with increasing concentrations of a soluble form of
cholesterol, 25-hydroxycholesterol (25HC) (Figure 5C, and S5B). Together, these
results show that CBD sensitizes cells to apoptosis when challenged with excess
cholesterol, either from endogenously synthesized or exogenous pools.
We next determined the impact of perturbations to cholesterol transport and storage on
CBD-induced apoptosis. We measured apoptosis in SK-N-BE(2) and HEK293T cells
treated with CBD and sublethal doses of 25HC (15 µg/ml), in combination with a
cholesterol transport inhibitor (NPC1 inhibitor U18666A, 10 µM), or an inhibitor of acyl-
coenzyme A cholesterol O-acyltransferase (ACAT), an enzyme required for
esterification and intracellular storage of cholesterol (VULM 1457, 5 µM). Both
compounds sensitized cells to apoptosis when CBD was present, which was more
pronounced when cells were also challenged with 25HC (Figure 5D and 5E). VULM
and U18666A treatment alone did not lead to increases in apoptosis. These results
Figure 5. CBD induced apoptosis is rescued by inhibitors of cholesterol synthesis and
increased by inhibitors of cholesterol transport and storage. (A) CBD was assessed for
cytotoxicity in SK-N-BE(2) cells across increasing doses of CBD at 24, 48 and 72 hours by CellTiter-
Glo luminescent assay. (B-E) SK-N-BE(2) or HEK293T cells were assessed for apoptosis at 24 hours
using live cell microscopy using a resazurin based fluorometric cell viability stain. Cells were treated
with 10 µM atorvastatin and exposed to increasing doses of CBD in B. 20 µM CBD and exposure to
increasing doses of 25-OH cholesterol in C, and combinations of 20 µM CBD, 10 µM U18666A, 5µM
VULM and 15µg/ ml 25- OH Cholesterol at 48 hours in D-E. Apoptosis displayed in a heatmap for each
condition. (See also Figure S5) (F) Live cell confocal microscopy of SK-N-BE(2) with NBD-Cholesterol
(Green) and a lysosomal dye Lyso-T (Red). Cholesterol subcellular distribution was examined upon
exposure of cells to 20 µM CBD and/ or 10 µM U18666A. Scale bar: 3 µM
indicate that interfering with cholesterol transport or cholesterol storage sensitize cells
treated with CBD to apoptosis.
One possible explanation for why CBD sensitizes cells to inhibitors of cholesterol
trafficking and storage is that CBD increases the flux of cholesterol transport from the
plasma membrane through the endosomal-lysosomal pathway. In support of this
hypothesis, we observed increases in apolipoproteins B and E (APOB, APOE)
abundance in the cytosol-enriched fraction. APOB/E are lipoprotein components of
cholesterol-containing low density lipoprotein particles required for cellular uptake of
cholesterol (Figure 3D). When cholesterol import through low density lipoprotein
receptors (LDLR) is activated, and 25HC is supplied in excess, the inability to efficiently
store cholesterol may cause subcellular accumulation of cholesterol in organelles that
normally maintain low cholesterol levels. To examine this possibility, we visualized
lysosomes (lysotracker dye) and cholesterol (NBD-cholesterol) in vehicle and CBD
treated SK-N-BE(2) cells with live cell confocal microscopy.
In cells treated with either CBD or U18666A alone, puncta stained with lysotracker and
NBD-cholesterol showed distinct spatial separation within cells. In contrast, co-
treatment with CBD and U18666A led to formation of enlarged, membranous organelles
co-stained with lysotracker and NBD-cholesterol, with a central unstained lumen-like
area (Figure 5F). Morphologically similar structures have been reported in models of
Niemann-Pick type C (NPC) disease, in which cholesterol accumulates in enlarged
lamellar inclusions with components of lysosomes and endosomes, leading to a toxic
cycle of enhanced cholesterol synthesis and intracellular accumulation (Demais et al.,
2016; Höglinger et al., 2019). Sensitization of cells by CBD to drugs that disrupt
intracellular trafficking of cholesterol, combined with our observations that CBD
increases endogenous cholesterol biosynthesis and stimulates cholesterol esterification
(Figures 5D,E and 4B,C), support a model where CBD increases transport of plasma
membrane cholesterol through the endosomal-lysosomal pathway to intracellular
compartments where it is esterified and sequestered, escaping ER-resident cholesterol
sensing machinery (Cheng et al., 1995; Lange et al., 1999).
CBD incorporates into membrane compartments altering cholesterol orientation
and lateral diffusion
Subcellular fractionation of SK-N-BE(2) cells treated with CBD for 24 hours showed that
CBD is concentrated primarily at the plasma membrane, with lower levels detected in
ER and nuclear membranes (Figure 6A). CBD accumulation in the plasma membrane,
and our evidence that CBD activates the internalization of plasma membrane
cholesterol, suggests that CBD alters cholesterol orientation and availability within the
plasma membrane. To measure the effect of CBD on cholesterol chemical activity, we
measured the enzymatic oxidation rate of cholesterol to 5-cholesten-3-one by
cholesterol oxidase in small unilamellar vesicles (SUVs). Cholesterol oxidase has been
shown to sense alterations of lipid bilayer structure and cholesterol orientation (Ahn and
Sampson, 2004), and could reveal CBD-dependent alterations in cholesterol orientation
in membranes. Titration of CBD into cholesterol-containing SUVs increased the initial
reaction rate of cholesterol oxidase in a manner proportional to CBD concentration
(Figure 6B and S6A).
Figure 6. CBD incorporates into membranes, increases cholesterol chemical activity, and
reduces lateral diffusion of cholesterol. (A) Ethanol extracts of subcellular fractions of SK-N-BE(2)
cells exposed to 0, 20 and 40 µM CBD for 24 hr were analyzed for CBD using LC-MS. (B) Synthetic
small unilamellar vesicles (SUVs-molar ratio: phosphatidylcholine:cholesterol:NBD cholesterol: 78:20:2
(n/n%) were used as a source of cholesterol in a fluorogenic cholesterol oxidase reaction to determine
the effect of CBD on initial reaction rate. (C) Identical experiments were performed with cholesterol
complexed to methyl beta cyclodextrin (MBCD), without SUVs present (D) Synthetic membrane
monolayers containing NBD-cholesterol were adsorbed to borosilicate glass and used in fluorescent
recovery after photobleaching (FRAP) experiments following exposure to either CBD (60 µM) and/or
DHA (20 µM). Scale bar is 2.5 µm. Quantified fluorescence recovery after photobleaching is displayed
in (E) and (F) (n=3).
The dose-dependent increase in cholesterol oxidase activity by CBD requires
cholesterol localized in membranes, as freely soluble 25HC (Figure S6B), and soluble
complexes of cholesterol and methyl beta cyclodextrin (MBCD), showed no dose
dependence on CBD (Figure 6C). We repeated this experiment in a complex
membrane environment using vesicles derived from ER membranes, and again
observed a concentration dependent increase in cholesterol oxidase activity in response
to CBD (Figure S6C). Together, these data provide evidence that CBD incorporates
into membranes and alters cholesterol accessibility, likely by altering cholesterol
orientation within the membrane to make the hydroxyl moiety more solvent accessible.
The ability of cholesterol oxidase assays to reveal alterations in lipid order has been
previously reported in studies noting that cholesterol oxidase can preferentially target
specialized ordered lipid domains known as caveolae (Ortegren et al., 2004; Smart et
al., 1994). Our results showing that CBD affects cholesterol activity in both synthetic
and cell derived ER membranes implies that CBD may contribute to increased lipid
order. A hallmark of increased lipid order is a decrease in lateral diffusion of lipids
(Ferreri, 2005; Lindblom and Orädd, 2009). We measured the effect of CBD on the
lateral diffusion of fluorescently labelled cholesterol (NBD-cholesterol) in synthetic
membrane monolayers. SUVs containing 20% (n/n%) of cholesterol and 2% (n/n%)
NBD-cholesterol were deposited on glass-bottom multiwell imaging plates, followed by
ultra-sonification. Recovery kinetics of fluorescent cholesterol were monitored in the
presence of vehicle or CBD using fluorescence recovery after photobleaching (FRAP)
(Figure 6D). CBD significantly reduced the recovery of fluorescence in the
photobleached monolayer area relative to vehicle control (Figure 6D and 6E),
suggesting that CBD slows the lateral diffusion of fluorescent cholesterol. This effect of
CBD on lateral diffusion could be rescued with simultaneous treatment of the
docosahexaenoic acid (DHA), a known disrupter of lipid order (Figures 6D and 6F).
Our FRAP experiments demonstrate that DHA and CBD have opposing effects on the
lateral diffusion of fluorescently labelled cholesterol in synthetic membranes. Although
this implies that CBD increases lipid order and that DHA decreases lipid order, it
remains unclear how these biophysical effects of CBD and DHA on cholesterol impact
cellular physiology. Esterification of DHA into membrane phospholipids results in
remodeling of sphingolipid/ cholesterol-enriched lipid rafts, a known hub for apoptosis
signaling (George and Wu, 2012; Wassall et al., 2018). To determine whether CBD and
DHA also have opposing effects in a cellular context, we quantified the effect of CBD
and DHA on cholesterol-dependent apoptosis. DHA treatment induced apoptosis in both
HEK293T and SK-N-BE(2) cells in a dose dependent manner (Figure S6D and S6E),
consistent with previous studies (Geng et al., 2018; Serini et al., 2008; Shin et al., 2013;
Sun et al., 2013). Importantly, this DHA-induced apoptosis proved to be cholesterol
dependent, as simultaneous treatment with DHA and the cholesterol sequestering
agent, MBCD, delayed apoptosis in HEK293T cells, and fully rescued apoptosis in SK-
N-BE(2) cells (Figure S6D and S6E). Similarly, CBD treatment (6.25 µM) rescued the
apoptotic effects of DHA in both HEK293T and SK-N-BE(2) cells at 48 hours (Figure
S6D and S6E). These data indicate that CBD and DHA have opposing effects on
cellular membrane structure and induction of apoptosis, both of which are cholesterol-
dependent, but the connection between these two processes remain unclear. Previous
observations that CBD induces lipid raft coalescence in mouse microglial cells (Wu et
al., 2012) suggests that CBD may alter lipid raft structure in a manner that frees lipid raft
associated cholesterol to internalization mechanisms.
Consistent with CBD alterations in cholesterol orientation, we found that CBD sensitized
live cells to the chemical agent filipin. Filipin is a highly fluorescent probe known to bind
cholesterol and disrupt nearby lipid ordering, resulting in permeabilization of
membranes. Cells pretreated with 20 µM CBD for 24 hours were preferentially
permeabilized by filipin relative to vehicle control (Figure S6F), further supporting a
CBD-dependent alteration in plasma membrane structure and cholesterol orientation.
This data suggests that CBD either directly increases cholesterol availability to filipin, or
destabilizes the membrane, thereby contributing to the membrane disruption effects of
Although clinical and preclinical evidence point to CBD as a promising therapeutic
compound for epilepsy, the cellular targets that mediate its effects in humans remain
largely unknown. In this study, we found that CBD elicits a diverse and pleiotropic effect
on human cells, using an unbiased profiling strategy of transcriptomics, subcellular
proteomics, phosphoproteomics, metabolomics, lipidomics, isotope tracing and live cell
microscopy. Our data suggest that CBD integrates into cellular membranes and alters
cholesterol orientation within the phospholipid environment. Partitioning of CBD into
model membranes increased lipid order, decreased lateral diffusion, and altered
cholesterol availability to the enzyme cholesterol oxidase, and would predict that CBD
alters the biophysical properties of cellular membranes membrane, with consequent
effects on diverse membrane proteins and their downstream targets.
We found that CBD treatment led to increased cytosolic calcium within 2 hours in
human neuroblastoma and keratinocyte cells. AMPK activity followed the observed
increase in calcium, suggesting upstream activation of AMPK by the calcium-dependent
kinase CAMKKβ. Compromised ATP generation by mitochondrial respiration may
sustain AMPK activation after 24 hours, as suggested by our Seahorse analysis.
Increased AMPK activity, and increased phosphorylation of its substrate ACACA,
predicted reduced fatty acid synthesis and altered Acetyl-CoA metabolism, which was
confirmed by flux metabolomics. Upregulation of cholesterol biosynthesis on proteome,
transcriptome and metabolomic levels occured as early as 3 hours and was sustained
up to 72 hours. As acetyl-CoA is precursor of cholesterol, the increase in cholesterol
biosynthesis precursors are consistent with acetyl-CoA supporting this flux. Parallel to
upregulation of cholesterol biosynthesis, increased cholesterol ester import occurs
through the LDLR-endocytic pathway, resulting in increased transport of cholesterol
through the lysosome. Increased stress on cholesterol regulatory processes, combined
with compromised cellular energetics driven by CBD may contribute to increased
apoptosis in CBD-treated cells. We propose that cholesterol is the primary
pharmacological effector of CBD in our human cell systems, rather than any given
protein receptor or target.
CBD integrates into cellular membranes and increases lipid order
Despite multiple lines of preclinical evidence outlining a wide array of phenotypic effects
of CBD (De Filippis et al., 2011; Esposito et al., 2011; Hinz and Ramer, 2019; Kenyon et
al., 2018; Massi et al., 2013; Pan et al., 2009; Rajesh et al., 2010), the molecular targets
that underlie these effects remain unknown. Our in vitro data with model membranes
demonstrate that CBD causes increased accessibility of cholesterol to cholesterol
oxidase, presumably through tight packing of cholesterol, CBD and possibly
phosphatidylcholine (Figure 6B). This result is surprising because it suggests that CBD
can induce ordered domains in the complete absence of sphingolipids, which are
historically used in formulations of model membranes for studying lipid order (Schroeder
et al., 1994; Silvius, 2003). Nevertheless, we detected the same behavior of cholesterol
in CBD treated ER membranes derived from subcellular fractionation of living cells
(Figure S6C), which implies that this ordering effect can also occur in cellular
membranes. Consistent with the effect of CBD on cholesterol orientation and packing,
we demonstrate that fluorescent cholesterol displays diminished lateral diffusion in
response to CBD exposure to synthetic membranes, and that this effect opposes, and
can be reversed by, the polyunsaturated lipid DHA (Figure 6D-F). Since diminished
lateral diffusion of lipids has been previously shown to be a hallmark of increased lipid
order, this data represents a second line of evidence for CBD induced order in
membranes. Additionally, the opposing effect between CBD and DHA on model
membranes was recapitulated in living cells in an opposing effect on apoptosis (Figure
7 B, C), which strengthens the case that CBD can affect the lipid order in cholesterol
containing membranes in living cells.
CBD has previously been reported to induce apoptosis in murine primary microglial cells
in a manner that was rescued by MBCD (Wu et al., 2012). MBCD is a selective agent
for the depletion of cholesterol from membranes by binding cholesterol within its
hydrophobic core (Kilsdonk et al., 1995; Klein et al., 1995). In this model, CBD
treatment enriched the membrane with large regions of lipid ordered domains marked
by gangliosides and caveolin-1. This lipid raft coalescence was also rescued by MBCD
treatment, providing a consistent model when combined with our data for the cholesterol
dependency of CBD-induced apoptosis. Within cells, lipid ordered domains (also
referred to as lipid rafts) are 10-120 nm wide transient domains in membrane structures
where tight packing of cholesterol, sphingolipids and GPI-anchored proteins allow for
activation of signaling cascades via the increased proximity of specific proteins
(Lingwood and Simons, 2010). These ordered domains have been implicated in
generating receptor signaling, assembly of endocytic machinery and endocytosis, and
regulation of intracellular calcium. Our data suggest that direct alterations in cholesterol
orientation and packing could underlie previously observed alterations in lipid raft
stability and size. Further, disruption of these lipid rafts is consistent with downstream
changes in calcium regulation, as well as cell death signaling as observed by FRET
sensor and caspase apoptosis assays presented here (Figure 1B and 5A) (George and
Wu, 2012; Liu et al., 2006; Pani and Singh, 2009).
Concordance of multi-omic data points to CBD disruption of cholesterol
Integration of our transcriptomics, metabolomics, and proteomics data provided multiple
lines of evidence for the disruption of cellular cholesterol homeostasis by the uptake of
CBD into human cell lines. Multiple aspects of cholesterol regulation were dysregulated
by CBD: cholesterol biosynthesis (Figure 3D and 4B), transport (Figure 3D and 5F),
and storage (Figure 4C). All three -omics methods provided evidence for perturbed
cholesterol biosynthesis. For instance, transcriptomics and proteomics reported
transcriptional activation and protein accumulation of the rate limiting enzyme in the
biosynthetic pathway of cholesterol, HMGCR (Figure 3D). Increased HMGCR protein
production is a canonical response to decreased cholesterol levels in the ER, where
cholesterol is sensed through the SREBP-SCAP axis (Brown and Goldstein, 1997).
Consistent with this observation, we found that cholesterol precursors accumulate in
CBD-treated cells (Figure 3B, S3A), with a modest decrease in total cholesterol
(Figure 3B, S3A, S3B) and a large increase in cholesterol esters.
Paradoxically, we found that CBD triggered the upregulation of cholesterol biosynthesis
enzymes, despite evidence showing only modest changes in total cholesterol, and
increased levels of cholesterol esters, which would normally result in downregulation of
cholesterol biosynthesis by the ER-resident SREBP sensing machinery (Brown and
Goldstein, 1997). These results suggest that in the presence of CBD, the ER is unable
to accurately sense the abundance of cholesterol at the plasma membrane, and as a
result, generates unneeded cholesterol that is esterified due to excessive intracellular
accumulation. GRAM domain proteins that localize to plasma membrane- ER contact
sites, bind and transport specific lipids between the two membranes (Besprozvannaya
et al., 2018). Within this family, GRAMD1s sense and bind the ‘accessible’ pool of
cholesterol that is not currently complexed with other lipid species, and transports it to
the ER (Naito et al., 2019). The pools of cholesterol that are either ‘accessible’ or
‘inaccessible/sequestered’ are regulated by the domains they associate with, and are
frequently driven by sphingomyelin and phospholipid association (Das et al., 2014;
Lange et al., 2013; Sokolov and Radhakrishnan, 2010). CBD driven alterations of
cholesterol orientation and decreased lateral diffusion presented here, together with
previously reported CBD-dependent increases in lipid raft stability and size, provide
strong evidence that the pool of ‘sequestered cholesterol’ is increased in CBD treated
conditions. However, specific investigation of GRAMD1 sensing and partitioning of CBD
and cholesterol within the ER membrane will need to be investigated in future studies.
The effects on cholesterol homeostasis in therapeutic applications of CBD
Our study demonstrates a global profiling strategy to identify key mechanistic
components in the CBD response, classify those components into broad biological
processes, and identify CBD as an efficient modulator of lipid order from which the
downstream effects originate. Our findings that CBD leads to disruption in cholesterol
and lipid homeostasis has broad implications on the mechanistic underpinnings of the
clinical effects of CBD in a wide array of diseases. Transmembrane proteins known to
be regulated through lipid ordered domains have been implicated in many of the
diseases for which CBD has been proposed as a therapeutic. These include
inflammatory disorders, Alzheimer’s disease, cancer (Gianfrancesco et al., 2018; Hsu et
al., 2018; Mollinedo and Gajate, 2015; Pirmoradi et al., 2019; Staneva et al., 2018).
Many of the targets of CBD proposed to underlie its efficacy as an anti-convulsant are
membrane proteins, including TRPV1, GPR55, and adenosine transport proteins
(Bisogno et al., 2001; Liou et al., 2008; Ryberg et al., 2007). CBD inhibits ion currents
from many structurally diverse voltage-gated ion channels at similar micromolar
concentrations, and with a high degree of cooperativity, suggesting that CBD acts
indirectly on ion channels through perturbation of membrane structure (Ghovanloo et
al., 2018). Moving forward, it will be important to determine the role of lipid order and
cholesterol orientation in the mediation of CBD-induced effects in models of generalized
Importantly, our study suggests that not all CBD effects on cells are therapeutically
beneficial, and that high dose use of CBD may lead to cholesterol-dependent side
effects in certain cell types that rely on high levels of cholesterol synthesis or import. We
demonstrated that CBD dependent apoptosis is heavily dependent on the cholesterol
status of cells (Figure 5B, 5C, S5A and S5B). As most of the cholesterol in humans is
synthesized in hepatic cells, we predict that many of the side effects of heavy CBD
consumption may occur in the liver. Indeed, some clinical evidence of the adverse side
effects of CBD in the liver has begun to emerge. Long term CBD use is associated with
elevated aminotransferase activity (Gaston et al., 2017), a hallmark of liver injury
(Giannini et al., 2005). Our data predict that CBD use may interact adversely with
certain dietary behaviors that elevate blood cholesterol, as the combination of
cholesterol/ hydroxycholesterol and CBD is toxic to a cell line derived from skin (HaCaT
cells), brain (SK-N-BE(2) cells), and kidney (HEK293T cells) (Figures 5D, 5E, S5B).
Further, our results show that CBD in combination with U18666A disrupts cholesterol
trafficking through lysosomes, raising the question of whether CBD use might increase
the risk of toxicity in patients with Niemann-Pick Disease, which harbor mutations in
NPC1, the target of U18666A (Lu et al., 2015).
By combining unbiased multi-omic profiling and biosensor activity screening, we
identified an unexpected pleiotropy of CBD action in human cell lines. Pharmacological
pleiotropy is a well-known phenomenon associated with steroid drugs such as statins
(Davignon, 2004), which target cholesterol biosynthesis, and raises an important
question of whether CBD acts directly through perturbation of membrane structure and
cholesterol content, or through inhibition of proteins involved in cholesterol trafficking
and synthesis. Our data supports a model where CBD partitions rapidly into cellular
membranes, disrupting the distribution of cholesterol in lipid rafts, leading to
compromised barrier function of the ER and PM, and dysregulated sensing and
biosynthesis of cholesterol.
The authors acknowledge the BioFrontiers Computing Core at the University of
Colorado Boulder for providing High Performance Computing resources (NIH
1S10OD012300) supported by BioFrontiers IT. The imaging work was performed at the
BioFrontiers Institute Advanced Light Microscopy Core. The Molecular Devices
ImageXpress was supported by NIH grant 1S10RR026680-01A1. Laser scanning
confocal microscopy was supported by NIST-CU Cooperative Agreement award
number 70NANB15H226. This work was supported by a DARPA cooperative
agreement, 13-34-RTA-FP-007, to WMO, MHBS, XL, and AD.
Conceptualization: W.O., X.L., and D.A.C.; Formal Analysis: J.J., S.C. and S.E.G.;
Funding Acquisition: W.O., M.H.B.S and X.L. Investigation: X.L., W.O., D.A.C., S.E.G.,
C.E., T.N., K.W., K.A.B., H.S., Z.P., J.R.H., and A.R.; Methodology: X.L., W.O., D.A.C.,
C.E., J.J. and S.C.; Project Administration: W.O.; Resources (purified hemp-derived
cannabidiol): R.S.; Supervision: M.H.B.S., A.D., X.L. and W.O.; Visualization: W.O.,
S.E.G., D.A.C., T.N., J.J., S.C., C.E. and E.B.; Writing - Original Draft: S.E.G., D.A.C.
and W.O.; Writing - Review & Editing: S.E.G. and W.O.
Declaration of Interests:
XL, DC and WO are patent holders of PCT WO2019246632A1, and XL, DC, WO and
EB are patent holders of PCT WO2019118837A1. Both patents are related to this work.
Unrelated to the contents of this manuscript, AD and TN are founders of Omix
Technologies Inc, D.C. is the founder of Biooloomics, Inc, and RS is the founder of
Sievers Infinity LLC.
CBD was derived from domestically grown industrial hemp that was cultivated and
purified by Sievers Infinity, LLC, Colorado-owned corporation, registered with the
Colorado Department of Agriculture (CDA) to grow and cultivate industrial hemp (CDA #
69096). The purified hemp-derived material was characterized by mass spectrometry,
X-ray diffraction, differential scanning calorimetry, nuclear magnetic resonance (1H-
NMR and 13C-) spectroscopy, and HPLC-UV. The quantitative proton NMR results
indicate that the sample is >95% CBD and the HPLC results indicate that 12 other
commonly found cannabinoids (including THC) were less than the limit of detection of
Generation of FRET Biosensor cell lines
Stable transgenic biosensor-expressing cell lines were made in HaCaT and SK-N-BE(2)
cells as previously described (Chapnick et al., 2015). Briefly, biosensor gene-containing
plasmids were obtained through the addgene plasmid depository, and subcloned into
our Bsr2 parent plasmid (sequence available upon request). Each biosensor Bsr2
plasmid was co-transfected with a PB recombinase expressing vector (mPB) via
polymer based transfection using polyethyleneimine (PEI) (Polysciences, 25kD Linear).
Each stable transgenic cell line was selected for 7 days using 10 µg/ml Blasticidin S.
FRET biosensor profiling was conducted in multiplexed parallel live cell experiments
using 384 well imaging plates (Corning #3985) in an ImageXpress MicroXL high
throughput microscope. Filters used for FRET measurements were the following: FRET
excitation 438/24-25, dichroic 520LP, emission 542/27-25 (Semrock MOLE-0189); CFP
excitation 438/24-25, dichroic 458LP, emission 483/32-25 (Semrock CFP-2432B-NTE-
Zero). Time lapse microscopy images were collected, and FRET Ratio calculations for
each site in each well at each time were performed as the mean value from pixels
above threshold of background and flatfield corrected images, where each pixel value
represented the FRET channel intensity divided by the CFP channel intensity. This
method is described in more detail in our previous studies (Chapnick and Liu, 2014;
Chapnick et al., 2015). Calculation and data visualization was performed in MATLAB
using custom scripts that are available upon request.
EC50 estimation from FRET sensor dose responses
Dose responses at each time point were fit with the following fit function y = 1 / (1 +
np.exp(-k*(x-EC50))) using python's scipy.optimize.curve_fit package. Prior to fitting,
measurements were scaled between zero and one. R^2 goodness of fit was calculated
between the sigmoid fit and the median of the replicates (duplicates) for each
sensor/timepoint combination. Fits with EC50 estimates outside the dose range were
discarded. EC50 values were kept for fits that resulted in a R^2 GOF > 0.75. The
resulting distribution of EC50 values was somewhat bimodal resulting in a median EC50
of 8.48 μM across all sensors and times.
CBD, and vehicle treatments were prepared in quadruplicate (4 drug treated/ 4 vehicle
controls) at 3, 6, 12, and 24 hour timepoints. 500ng of Total RNA was used in Illumina
TruSEQ mRNA library prep protocol. Libraries were run on the Illumina HiSEQ4000 at
single read 150 bp length. Sequencing was performed on seven consecutive lanes.
Median read counts per lane were ~49,000 with a CV of ~7%. Starting with 228 fastq
files, each lane set was concatenated per condition. Run specifications were, 51 bp
reads, standard single-read, first-stranded. Alignment to the human genome (HG19),
was done using Tophat version 2.0.13. Two mismatches were allowed, and duplicate
mapping was permitted at up to two locations. Using Trimmomatic-0.36, base pairs
were trimmed if they dropped below a PHRED score of 15 within a sliding window of
4bp. Remaining reads shorter than 35 base pairs were removed. Illumina adapter
sequences were also clipped using Trimmomatic. Fastqc was used to verify data
integrity before and after trimming. Cufflinks/Cuffdiff 2.2.1 was used to obtain FPKM
normalized gene count and differential expression measurements at each time point
(Trapnell et al., 2010). One p/q-value was generated for each gene at each timepoint.
Genes with a q-value significance < 0.05, and absolute log2 fold change of 1 or greater,
for at least one time point, were retained for downstream analysis.
Proteomics Bioinformatics Workflow:
Protein quantification for time-series was performed with a Tandem Mass Tag (TMT)
isobarically labeled 11-plex multiplexing scheme. The 15-point time series for each
cellular fraction was split into three series, with every series containing 5 treatment and
matched control time point pairs, with 0 sec, 40 min, 3 hr, 12 hr, and 24 hr time points in
Series A; 10 min, 80 min, 6 hr, 15 hr, and 48 hr in Series B; and 20 min, 2 hr, 9 hr, 18
hr, and 72 hr time points in Series C. This separation was performed so that a protein
could be missing from one and/or two series due to stochastic effects in data-dependent
acquisition and the overall trend could still be inferred, though with reduced resolution.
The 11th label in each series was devoted to a global mix reference channel, which
would be present in all series for a given cellular fraction. The global mix is a cell-
fraction specific mixture that contains an equal portion from each time point sample.
This channel was the denominator in the intermediate ratiometric measurement for
differential expression for both drug-treated samples and time-matched controls. This
mixture channel was constructed so that every measurable protein observed at any time
point has a non-zero denominator when ratios are taken. When the differential
expression is compared between the drug-treated labeled samples and matched control
samples and expressed as a log2 ratio, the global mix reference channel cancels out.
The differential expression of each individual protein was determined using Bayesian
methods for isobaric labeled proteomics data (Jow et al., 2014). Briefly, all observed
peptides are mapped to a list of observed protein ID’s via Isoform Resolver (Meyer-
Arendt et al., 2011). The TMT 11-plex reporter ion spectrum peaks for each peptide
contributes to the inference of the differential expression of a protein and reporter ion
label. In this case, each reporter ion label represents a particular measured time point.
The label-to-label normalization is handled via a hierarchical model, which calculates
the bias inherent with each specific label by pooling differential expression estimates
from all proteins, changing and unchanging. The hierarchical models are solved
computationally with a Markov Chain Monte Carlo (MCMC) method, running chains in
parallel for faster results (Denwood, 2016). The MCMC returns a Gaussian posterior
probability distribution of log2 differential expression for each protein for each label. The
model initially fits the ratiometric differential expression for every treatment and matched
control relative to a global mix channel, and the reported drug-induced differential
expression is the difference (in log2 space) between the treated sample and the
matched control sample. Five MCMC chains were run independently for at least 500k
steps, and convergences for each run were verified via Gelman-Rubin convergence
variable < 1.05 (Gelman and Rubin, 1992).
The differential expression was calculated independently for all biological replicates so
protein-level variance from separate replicates could be examined and quantified in the
posterior distributions obtained from MCMC. For reporting a single differential
expression for a protein and label, the Bayesian updating procedure is used to produce
a single posterior distribution, from which a mean point estimate and 95% credible
interval are calculated. In some specific instances, labels represent technical rather than
biological replicates. In cases of technical replicates, the point estimate values were
averaged and the credible intervals extents were treated as errors and added in
quadrature. With this procedure, technical replicates contribute a single probability
distribution to any further Bayesian updating.
For every cellular fraction and time point, then, there are between 3 and 6 biological
replicates, and the number of replicates represented in the drug treated samples and
the matched control samples are not necessarily the same. The effect size (Cohen’s d)
was calculated between the posterior probability distributions of the drug treated and
matched control samples as a standardized measure to determine if there was a drug
effect. Statistical power analysis was performed to show that, with significance criteria α
= 0.05 and the statistical power requirement (1-β) = 0.8, the appropriate effect size
threshold should be d > (1.50, 1.75, 2.25, 3.38) for proteins observed within 6, 5, 4, or 3
replicates, respectively. A protein was selected for further consideration if it showed
differential expression greater than this threshold for any given time point.
The Bioconductor Edge package (Leek et al., 2006) (DOI:10.18129/B9.bioc.edge)
version 2.8.0 was used for time course differential analysis. Many proteins were not
present for all replicates and/or plexes, so Edge was run sequentially to generate p-
values for each case. For instance, in the soluble fraction, there were 273 proteins that
were only present in two replicates. These were run through Edge separately from the
other 1957 proteins that were observed in three replicates. The resulting time series p-
values were combined into a list and FDR corrected using Benjamini-Hochberg multiple
hypothesis correction (Benjamini and Hochberg, 1995).
Proteomics Network Analysis
Of the significantly changing proteins, correlation networks were generated for each
subcellular fraction. Networks were created from the ethanol (vehicle) treated samples,
as well as for the CBD treated samples. Network edge values were assigned using
Spearman correlation coefficients between all proteins (vertices) for a given
replicate. For each pair of proteins, 2*N edge values were generated, where N is the
number of available replicate measurements for that protein. An independent t-test was
used between basal replicate edge values and treatment edge values to evaluate what
edges were significantly changed due to CBD treatment. Edges with -log10(p-value) > 2
(p < 1%) were retained. Python graph-tool package was used to generate a stochastic
block-model representation of the resulting network, which clusters nodes based on
network connectivity similarity.
Combined Heatmap Criteria: All protein IDs with edge adjusted p-values less than 1%
were merged with gene IDs from RNAseq with edge adjusted p-value < 1% and
minimum absolute log2 fold change > 0.5. Merged list was used as input for Enrichr
(Kuleshov et al., 2016) to get a table of GO terms (go_biological_processes_2017). GO
terms were reduced using REVIGO with “medium” size setting: Terms with
dispensability score less than 0.1 and q-value < 5% were kept. Merged IDs from
remaining GO ontologies were clustered and plotted in heatmap by relative expression
in CBD treated condition compared to vehicle control at each time point starting at 3
Subcellular Fractionation in Proteomics
For each sample, a 10 cm Petri Dish containing 106 SK-N-BE(2) cells was harvested
and washed three times with 10 ml of 20°C PBS. All PBS was removed by aspiration
and plates were frozen using liquid nitrogen and stored at -80°C overnight. Each plate
was thawed on ice and 400 µl Tween20 Buffer (1x PBS, 0.1 % Tween20, 5 mM EDTA,
30 mM NaF, 1 mM NaVo4, 100 µM Leupeptin, 2.5 µM Pepstatin A) and scraped
thoroughly using a standard cell scraper. The resulting lysate was homogenized with a
200 µl pipette and transferred to 1.7 mL Eppendorf tube on ice. Lysate tubes were
incubated for 30 min at 4°C rotating end-over-end. After rotation, tubes were centrifuged
for 10 min at 4°C (16,100 rcf). All supernatant was transferred into new labeled 1.7mL
Eppendorf. This tube contains insoluble buoyant plasma membrane and cytosol. The
leftover pellet is the ‘Membrane’ fraction and is enriched in nuclei. 40µL of 1 M NaOAc
was added to the supernatants, which immediately were exposed to centrifugation for
10 min at 4°C (16,100 rcf). All supernatant was transferred into new labeled 1.7mL
Eppendorf. This is the ‘Soluble’ fraction. The pellet was resuspended in 400 µl 20 °C
SDS buffer. This is ‘Insoluble #2’ fraction. All fraction containing tubes were filled
completely with -20 °C Acetone and stored overnight in -20 °C. Each tube was exposed
to centrifugation for 10 min at 4°C (16,100 rcf) and supernatants were aspirated and
discarded, while pellets were allowed pellets to air dry 10 min 20 °C. The pellets then
proceeded to the FASP procedure.
Quantitative Subcellular Proteomics
Precipitated and dried subcellular protein extracts were solubilized with 4% (w/v)
sodium dodecyl sulfate (SDS), 10mM Tris(2-carboxyethyl)phosphine (TCEP), 40mM
chloroacetamide with 100mM Tris base pH 8.5. SDS lysates were boiled at 95C for 10
minutes and then 10 cycles in a Bioruptor Pico (Diagenode) of 30 seconds on and 30
second off per cycle, or until protein pellets were completely dissolved. Samples were
then cleared at 21,130 x g for 10 minutes at 20⁰C, then digested into tryptic peptides
using the filter-aided sample preparation (FASP) method (Wiśniewski, 2016). Briefly,
SDS lysate samples were diluted 10-fold with 8M Urea, 0.1M Tris pH8.5 and loaded
onto an Amicon Ultra 0.5mL 30kD NMWL cutoff (Millipore) ultrafiltration device.
Samples were washed in the filters three time with 8M Urea, 0.1M Tris pH8.5, and again
three times with 0.1M Tris pH8.5. Endoproteinase Lys-C (Wako) was added and
incubated 2 hours rocking at room temperature, followed by trypsin (Pierce) which was
incubated overnight rocking at room temperature. Tryptic peptides were eluted via
centrifugation for 10 minutes at 10,000 x g, and desalted using an Oasis HLB cartridge
(Waters) according to the manufacture instructions.
High pH C18 fractionation of TMT labeled peptides
Dried 10-plexed samples were then suspended in 20uL 3% (v/v) acetonitrile (ACN) and
0.1% (v/v) trifluoroacetic acid (TFA) and loaded onto a custom fabricated reverse phase
C18 column (0.5 x 200mm C18, 1.8um 120Å Uchrom (nanoLCMS Solutions)
maintained at 25⁰C and running 15uL/min with buffer A, 10mM ammonium formate, pH
10 and buffer B, 10mM ammonium formate, pH10 in 80% (v/v) ACN with a Waters M-
class UPLC (Waters). Peptides were separated by gradient elution from 3% B to 50% B
in 25 minutes, then from 50% B to 100% B in 5 minutes. Fractions were collected in
seven rounds of concatenation for 30 s per fraction, then combined for a final of six high
pH C18 fractions. Samples were dried and stored at -80⁰C until ready for LC/MS
Liquid Chromatography/Mass spectrometry analysis
Samples were suspended in 3% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid and
direct injected onto a 1.7um, 130Å C18, 75um X 250mm M-class column (Waters), with
a Waters M-class UPLC or a nanoLC1000 (Thermo Scientific). Tryptic peptides were
gradient-eluted at 300nL/minute, from 3% acetonitrile to 20% acetonitrile in 100 minutes
into an Orbitrap Fusion mass spectrometer (Thermo Scientific). Precursor mass
spectrums (MS1) were acquired at 120,000 resolution from 380-1500 m/z with an
automatic gain control (AGC) target of 2.0E5 and a maximum injection time of 50ms.
Dynamics exclusion was set for 15 seconds with a mass tolerance of +/- 10 ppm.
Quadrupole isolation for MS2 scans was 1.6 Da sequencing the most intense ions using
Top Speed for a 3 second cycle time. All MS2 sequencing was performed using
collision induced dissociation (CID) at 35% collision energy and scanned in the linear
ion trap. An AGC target of 1.0E4 and 35 second maximum injection time was used.
Selected-precursor selections of MS2 scans was used to isolate the five most intense
MS2 fragment ions per scan to fragment at 65% collision energy using higher energy
collision dissociation (HCD) with liberated TMT reporter ions scanned in the orbitrap at
60,000 resolution (FWHM). An AGC target of 1.0E5 and 240 second maximum injection
time was used for all MS3 scans. All raw files were converted to mzML files and
searched against the Uniprot Human database downloaded April 1, 2015 using Mascot
v2.5 with cysteine carbamidomethylation as a fixed modification, methionine oxidation,
and protein N-terminal acetylation were searched as variable modifications. Peptide
mass tolerance was 20ppm for MS1 and 0.5mDa for MS2. All peptides were
thresholded at a 1% false discovery rate (FDR).
Sample preparation and phosphopeptide enrichment
SK-N-BE(2) cells were cultured in SILAC media either with Lys8 and Arg10 (Heavy) or
Lys0 and Arg0 (Light). Two biological replicates of near confluent Heavy cells and two
replicates of near confluent Light cells were treated with 20uM CBD for 10 minutes (4
replicates), 1 hour (4 replicates) and 3 hours (4 replicates) for phosphoproteomics
analysis. Cells were harvested in 4% (w/v) SDS, 100mM Tris, pH 8.5 and boiled at
95⁰C for 5 minutes. Samples were reduced with 10mM TCEP and alkylated with 50mM
chloroacetamide, then digested using the FASP protocol, with the following
modifications: an Amicon Ultra 0.5mL 10kD NMWL cutoff (Millipore) ultrafiltration device
was used rather than a 30kD NMWL cutoff. Tryptic peptides were cleaned a Water HLB
Oasis cartridge (Waters) and eluted with 65% (v/v) ACN, 1% TFA. Glutamic acid was
added to 140mM and TiO2 (Titanshere, GL Sciences) was added at a ratio of 10mg
TiO:1 mg tryptic peptides and incubated for 15 minutes at ambient. The
phosphopeptide-bound TiO2 beads were washed with 65% (v/v) ACN, 0.5% TFA and
again with 65% (v/v) ACN, 0.1% TFA, then transferred to a 200uL C8 Stage Tip
(Thermo Scientific). Phosphopeptides were eluted with 65% (v/v) ACN, 1% (v/v)
ammonium hydroxide and lyophilized dry.
High pH C18 fractionation of enriched phosphopeptides
Enriched phosphopeptide samples were then suspended in 20uL 3% (v/v) acetonitrile
(ACN) and 0.1% (v/v) trifluoroacetic acid (TFA) and loaded onto a custom fabricated
reverse phase C18 column (0.5 x 200mm C18, 1.8um 120Å Uchrom (nanoLCMS
Solutions) maintained at 25⁰C and running 15uL/min with buffer A, 10mM ammonium
formate, pH 10 and buffer B, 10mM ammonium formate, pH10 in 80% (v/v) ACN with a
Waters M-class UPLC (Waters). Peptides were separated by gradient elution from 3% B
to 50% B in 25 minutes, then from 50% B to 100% B in 5 minutes. Fractions were
collected in seven rounds of concatenation for 30 sec per fraction for a final of twelve
high pH C18 fractions. Samples were dried and stored at -80⁰C until analysis.
Liquid Chromatography/Mass spectrometry analysis of phosphopeptide fractions
Samples were suspended in 3% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid and
direct injected onto a 1.7um, 130Å C18, 75um X 250mm M-class column (Waters), with
a Waters M-class UPLC. Tryptic peptides were gradient eluted at 300nL/minute, from
3% acetonitrile to 20% acetonitrile in 100 minutes into an Orbitrap Fusion mass
spectrometer (Thermo Scientific). Precursor mass spectrums (MS1) were acquired at
120,000 resolution from 380-1500 m/z with an AGC target of 2.0E5 and a maximum
injection time of 50ms. Dynamics exclusion was set for 20 seconds with a mass
tolerance of +/- 10 ppm. Isolation for MS2 scans was 1.6 Da using the quadrupole, and
the most intense ions were sequenced using Top Speed for a 3 second cycle time. All
MS2 sequencing was performed using higher energy collision dissociation (HCD) at
35% collision energy and scanned in the linear ion trap. An AGC target of 1.0E4 and 35
second maximum injection time was used. Rawfiles were searched against the Uniprot
human database using Maxquant with cysteine carbamidomethylation as a fixed
modification. Methionine oxidation, protein N-terminal acetylation, and phosphorylation
of serine, threonine and tyrosine were searched as variable modifications. All peptides
and proteins were thresholded at a 1% false discovery rate (FDR).
Bulk Metabolomics Sample Preparation
Cultured cells were harvested, washed with PBS, flash frozen, and stored at -80C until
analysis. Prior to LC-MS analysis, samples were placed on ice and re-suspended with
methanol:acetonitrile:water (5:3:2, v/v/v) at a concentration of 2 million cells per ml.
Suspensions were vortexed continuously for 30 min at 4°C. Insoluble material was
removed by centrifugation at 10,000 g for 10 min at 4°C and supernatants were isolated
for metabolomics analysis by UHPLC-MS. This method was used for cholesterol
precursors and free headgroups.
UHPLC-MS analysis for Bulk Metabolomics
Analyses were performed as previously published (Nemkov et al., 2017, 2019). Briefly,
the analytical platform employs a Vanquish UHPLC system (Thermo Fisher Scientific,
San Jose, CA, USA) coupled online to a Q Exactive mass spectrometer (Thermo Fisher
Scientific, San Jose, CA, USA). Samples were resolved over a Kinetex C18 column,
2.1 x 150 mm, 1.7 µm particle size (Phenomenex, Torrance, CA, USA) equipped with a
guard column (SecurityGuardTM Ultracartridge – UHPLC C18 for 2.1 mm ID Columns –
AJO-8782 – Phenomenex, Torrance, CA, USA) (A) of water and 0.1% formic acid and a
mobile phase (B) of acetonitrile and 0.1% formic acid for positive ion polarity mode, and
an aqueous phase (A) of water:acetonitrile (95:5) with 1 mM ammonium acetate and a
mobile phase (B) of acetonitrile:water (95:5) with 1 mM ammonium acetate for negative
ion polarity mode. Samples were eluted from the column using either an isocratic elution
of 5% B flowed at 250 µl/min and 25ºC or a gradient from 5% to 95% B over 1 minute,
followed by an isocratic hold at 95% B for 2 minutes, flowed at 400 µl/min and 30ºC.
The Q Exactive mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) was
operated independently in positive or negative ion mode, scanning in Full MS mode (2
μscans) from 60 to 900 m/z at 70,000 resolution, with 4 kV spray voltage, 15 sheath
gas, 5 auxiliary gas. Calibration was performed prior to analysis using the PierceTM
Positive and Negative Ion Calibration Solutions (Thermo Fisher Scientific). Acquired
data was then converted from raw to .mzXML file format using Mass Matrix (Cleveland,
OH, USA). Metabolite assignments, isotopologue distributions, and correction for
expected natural abundances of deuterium, 13C, and 15N isotopes were performed using
MAVEN (Princeton, NJ, USA).(Clasquin et al., 2012) Graphs, heat maps and statistical
analyses (either T-Test or ANOVA), metabolic pathway analysis, PLS-DA and
hierarchical clustering was performed using the MetaboAnalyst package
(www.metaboanalyst.com)(Chong et al., 2018).
Lipidomics Sample Preparation
Extraction of cholesterol, precursors, free fatty acids, cholesteryl esters, and phospho
lipids were performed in the following manner. SK-N-BE(2) cells in 10 cm dishes were
washed with 10 mL PBS twice and then cells were scraped and pelleted at 400 rcf for 2
minutes. Cell pellets were resuspended in 100% methanol at 4°C and sonicate at 70%
power in 10 pulses, 5 seconds on/5 seconds off. The resulting lysate was rotated for 60
minutes at room temperature, followed by centrifugation for 20 min at 4°C (16,100 rcf).
Subcellular fractionation of organelles from intact SK-N-BE(2) cells was done in the
following manner to assess subcellular CBD distribution. Cells in 10 cm culture dishes
were harvest by washing twice with 10 mL PBS at room temperature, followed by
trypsinization using a cell culture grade Trypsin/EDTA solution (ThermoFisher).
Trypsinized cells were quenched by addition of 2 mL 10% FBS containing DMEM and
cells were pelleted by centrifugation for 2 min at 4°C (200 rcf). Cell pellets were wash
one time with 10 mL PBS, and resuspended in 1 mL Tween20 Buffer (1x PBS, 0.05 %
Tween20, 5 mM EDTA). This lysate was subjected to mechanical disruption using a
1mL glass Dounce homogenizer, 10 full passes at 4°C. Nuclei was pelleted from
homogenate by centrifugation for 5 min at 4°C (2,000 rcf). Supernatant was separated
and insoluble ER membranes were pelleted by centrifugation for 10 min at 4°C (4,000
rcf). Supernatant was separated and insoluble plasma membranes were pelleted by
centrifugation for 10 min at 4°C (16,000 rcf). Extraction of all fractions was done in
100% methanol for 2 hours at room temperature and rotation end-over-end, followed by
removal of insoluble material by centrifugation for 20 min at 20°C (16,100 rcf).
UHPLC-MS analysis for Lipidomics
Samples were analyzed as published (Reisz et al., 2019). Briefly, analytes were
resolved over an ACQUITY HSS T3 column (2.1 x 150 mm, 1.8 µm particle size
(Waters, MA, USA) using an aqueous phase (A) of 25% acetonitrile and 5 mM
ammonium acetate and a mobile phase (B) of 90% isopropanol, 10% acetonitrile and 5
mM ammonium acetate. The column was equilibrated at 30% B, and upon injection of
10 μL of extract, samples were eluted from the column using the solvent gradient: 0-9
min 30-100% B and 0.325 mL/min; hold at 100% B for 3 min at 0.3 mL/min, and then
decrease to 30% over 0.5 min at 0.4 ml/min, followed by a re-equilibration hold at 30%
B for 2.5 minutes at 0.4 ml/min. The Q-Exactive mass spectrometer (Thermo Fisher)
was operated in positive and negative ion mode using electrospray ionization, scanning
in Full MS mode (2 μscans) from 150 to 1500 m/z at 70,000 resolution, with 4 kV spray
voltage, 45 shealth gas, 15 auxiliary gas. When required, dd-MS2 was performed at
17,500 resolution, AGC target = 1e5, maximum IT = 50 ms, and stepped NCE of 25, 35
for positive mode, and 20, 24, and 28 for negative mode. Calibration was performed
prior to analysis using the PierceTM Positive and Negative Ion Calibration Solutions
(Thermo Fisher). Acquired data was then converted from .raw to .mzXML file format
using Mass Matrix (Cleveland, OH, USA). Samples were analyzed in randomized order
with a technical mixture injected incrementally to qualify instrument performance. This
technical mixture was also injected three times per polarity mode and analyzed with the
parameters above, except CID fragmentation was included for unknown compound
identification. Metabolite assignments were made based on accurate intact mass (sub 5
ppm), isotope distributions, and relative retention times, and comparison to analytical
standards in the SPLASH Lipidomix Mass Spec Standard (Avanti Polar Lipids) using
MAVEN (Princeton, NJ, USA). Discovery mode analysis was performed with standard
workflows using Compound Discoverer and Lipid Search 4.0 (Thermo Fisher Scientific,
San Jose, CA).
Confocal Microscopy of Cholesterol and Lysosomes
SK-N-BE(2) cells were seeded into fibronectin coated glass bottom 96 well plates
(Matriplate) at a cell density of 40,000 cells/well using low background imaging media
(FluoroBrite DMEM with all supplements described, above). At the time of seeding,
lysotracker Deep Red (ThermoFisher) was added at a 1000x dilution and NBD-
cholesterol (ThermoFisher) was added at a final concentration of 10 µg/mL. After 24
hrs, CBD or ethanol vehicle was added to a final concentration of 20 µM and incubated
for an additional 24 hrs prior to imaging using a Nikon A1R laser scanning confocal
microscope for acquisition with the FITC and TRITC channels. In experiments using
U18666A, a final concentration of 10 µg/ml was used and was added simultaneously
Assaying Cell Viability and Apoptosis
Cell viability for SK-N-BE(2) cells was conducted using a fluorometric cell viability assay
using Resazurin (PromoKine) according to the manufacturer’s instructions.
Measurement of percent apoptotic cells was done in 384 well imaging plates (Corning
#3985) seeded with 2,000 cells/well and stained with Hoescht 33258 (1µg/mL) and
CellEvent Caspase-3/7 Green Detection Reagent (ThermoFisher) at a dilution of 1000x.
Dyes were added at the time of seeding, 18-24 hours prior to performing experiments.
For experiments using atorvastatin, atorvastatin was added 24 hrs prior to addition of
CBD. For experiments involving 25-hydroxy cholesterol, U18666A, and VULM 1457,
inhibitors were added simultaneously with CBD. Experiments were performed using an
ImageXpress MicroXL microscope and a 10x objective, where images were acquired for
each well at the indicated time-points using DAPI and FITC filter sets. Using MATLAB,
images were processed with custom written scripts (available upon request) that
perform flatfield & background correction, identification of all cells (DAPI channel) using
standard threshold above background techniques, and identification of apoptotic cells
using a similar method in the FITC channel. Percent apoptotic cells was calculated from
the sum of apoptotic cell pixels divided by the sum of all cell pixels for each field of view.
Error displayed is the standard deviation from between 2 and 4 biological replicates.
Fluorometric Cholesterol Oxidase Experiments and SUV preparation
SUVs were prepared by dissolving 10 mg L-α-Phosphatidylcholine (Sigma P3556) in
100 µL chloroform in a glass vial, followed by removal of solvent under vacuum at room
temperature for 1 hour. For experiments using cholesterol containing SUVs, 0.74 mg of
cholesterol (Sigma C8667) was mixed with 10 mg L-α-Phosphatidylcholine prior to
removal of chloroform solvent. Following solvent removal, 100 µL PBS was added and a
microtip sonicator was inserted to perform sonication at 70% power, 10 pulses, 5
seconds on/off at room temperature. SUVs in suspension were brought to a volume of 1
mL with addition of PBS. The resulting SUVs in suspension were used at a dilution of
100-fold in subsequent cholesterol oxidase reactions. Cholesterol oxidase reactions
were performed using reagents from the Amplex Red Cholesterol Assay Kit
(ThermoFisher A12216), where each reaction was performed in 50 µL volumes using:
0.5 µL SUVs solution, 0.05 µL cholesterol oxidase solution, 0.05 µL HRP solution, 0.05
µL Amplex Red/DMSO made according to manufacturer’s instructions, and the
indicated CBD concentrations in PBS. Reaction volume was brought to 50 µL using
PBS. In cases where SUVs were not used, either 1 µg/reaction 25-OH cholesterol or 5
µg/reaction MBCD:Cholesterol (1:2) was substituted for SUVs. MBCD:Cholesterol was
prepared as previously described (Widenmaier et al., 2017). Cholesterol oxidase
reactions were performed in Corning 384 well optical imaging plates (#3985) in an
ImageXpress MicroXL widefield fluorescence microscope using the TRITC filter sets,
where 1 ms exposure time images were taken of each well 20 µm above the well
bottom every 10 minutes for 5 hours at 37 °C, using a 10x objective. Images were
flatfield corrected and the sum of fluorescence intensity across all 540x540 pixels was
calculated using custom MATLAB scripts that are available upon request. Product
formation of Amplex red was found to be linear within 0-1 hours, and data between t=0
and t=1 hours was used to calculate the average rate of increase in TRITC fluorescence
using Microsoft Excel. Displayed error bars represent the standard deviation of three or
more replicate reactions.
FRAP Experiments Using Synthetic Membranes
The formulation and techniques to create SUVs were repeated with addition of 108 µg
of 22-NBD cholesterol to L-α-Phosphatidylcholine and cholesterol prior to the solvent
removal step described for preparation of SUVs. A 1:5 dilution of NBD-cholesterol
containing SUV suspension:PBS was added to each well of glass bottom 96 well plates
(MatriPlate MGB096-1-2-LG-L) such that each well contains 100 µl of diluted SUV
suspension. 96 well plates were exposed to centrifugation for 20 minutes at 2000 rcf
using a swinging bucket rotor at room temperature. A microtip sonicator was inserted
into each well to perform sonication at 20% power, 20 pulses, 2 seconds on/off at room
temperature. The contents of each well was washed three times with 150 µl of PBS, and
subsequent experiments were performed with 250 µl PBS containing ethanol vehicle, 20
µM CBD and/or 20 µM DHA. CBD and DHA were incubated in wells for 1 hr at room
temperature prior to imaging and FRAP experiments using a Nikon A1R microscope.
Photobleaching was performed using Nikon Elements software with the following
parameters: framerate 250 ms, 100% power 488 laser for photobleaching for 250 ms,
and optical settings for FITC. Analysis was performed using ImageJ and Microsoft
Excel. All trends were normalized by division of mean intensity within the photobleached
region to a region of identical size remote from the photobleached region. Error bars
indicate the standard error of the mean from three replicates.
Seahorse Extracellular Flux Analysis
Oxygen consumption rate and extracellular acidification rate were measured using the
SeahorseXFe24 Extracellular Flux Analyzer and the Agilent Seahorse XF Cell Energy
Phenotype Test Kit. Cells were plated at 2x10^4 cells per well in XFe24 microplates.
Cells were treated with either 20 µM CBD or ethanol as a vehicle control either 24 hours
or 2 hours prior to assaying. The day of the assay cells were washed with an assay
medium containing 20 µM CBD or vehicle and placed at 37°C in a CO2 free incubator
for 1 hour. 1 µM Oligomycin and 1 µM FCCP were injected by the Seahorse analyzer as
oxygen consumption rate and extracellular acidification rate were measured per
All unique/stable reagents generated in this study are available from the lead contact,
William Old (email@example.com) with a completed Materials Transfer Agreement.
Data and Code Availability Statements
Proteomics and phosphoproteomics raw data is available at the MassiVE repository ID
MSV000085479 accessible at https://doi.org/doi:10.25345/C5571V
Source data for RNAseq experiments is accessible at GEO with the identifier
GSE151512 at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE151512
The code generated during this study is available at GitHub,
DOI:10.5281/zenodo.3861043, URL: https://github.com/CUOldLab/cbd-manuscript-
Supplemental Tables Summary
Table S1. Global Metabolomics Results, Related to Figure 1C and Supplementary
Table S2. Phosphoproteome Results, Related to Figure 2
Table S3. Cytosolic Proteome Results, Related to Figure 3
Table S4. Nuclear Proteome Results, Related to Figure 3
Table S5. Membrane Proteome Results, Related to Figure 3
Table S6. RNAseq Differential Analysis Results, Related to Figure 3D
Table S7. Lipidomics Results, Related to Figure 4
Ahn, K., and Sampson, N.S. (2004). Cholesterol Oxidase Senses Subtle Changes in
Lipid Bilayer Structure. Biochemistry 43, 827–836.
Ahrens, J., Demir, R., Leuwer, M., de la Roche, J., Krampfl, K., Foadi, N., Karst, M., and
Haeseler, G. (2009). The nonpsychotropic cannabinoid cannabidiol modulates and
directly activates alpha-1 and alpha-1-Beta glycine receptor function. Pharmacology 83,
Alharris, E., Singh, N.P., Nagarkatti, P.S., Nagarkatti, M., Alharris, E., Singh, N.P.,
Nagarkatti, P.S., and Nagarkatti, M. (2019). Role of miRNA in the regulation of
cannabidiol-mediated apoptosis in neuroblastoma cells. Oncotarget 10, 45–59.
Amsalem, M., Poilbout, C., Ferracci, G., Delmas, P., and Padilla, F. (2018). Membrane
cholesterol depletion as a trigger of Nav1.9 channel‐mediated inflammatory pain. EMBO
Bang, B., Gniadecki, R., and Gajkowska, B. (2005). Disruption of lipid rafts causes
apoptotic cell death in HaCaT keratinocytes. Exp. Dermatol. 14, 266–272.
Benjamini, Y., and Hochberg, Y. (1995). Controlling the False Discovery Rate: A
Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical
Society: Series B (Methodological) 57, 289–300.
Berger, L., Slein, M.W., Colowick, S.P., and Cori, C.F. (1946). ISOLATION OF
HEXOKINASE FROM BAKER’S YEAST. J. Gen. Physiol. 29, 379–391.
Besprozvannaya, M., Dickson, E., Li, H., Ginburg, K.S., Bers, D.M., Auwerx, J., and
Nunnari, J. (2018). GRAM domain proteins specialize functionally distinct ER-PM
contact sites in human cells. ELife 7, e31019.
Bisogno, T., Hanuš, L., De Petrocellis, L., Tchilibon, S., Ponde, D.E., Brandi, I., Moriello,
A.S., Davis, J.B., Mechoulam, R., and Di Marzo, V. (2001). Molecular targets for
cannabidiol and its synthetic analogues: Effect on vanilloid VR1 receptors and on the
cellular uptake and enzymatic hydrolysis of anandamide. British Journal of
Pharmacology 134, 845–852.
Blais, J.D., Filipenko, V., Bi, M., Harding, H.P., Ron, D., Koumenis, C., Wouters, B.G.,
and Bell, J.C. (2004). Activating transcription factor 4 is translationally regulated by
hypoxic stress. Mol. Cell. Biol. 24, 7469–7482.
Blum, B.C., Mousavi, F., and Emili, A. (2018). Single-platform ‘multi-omic’ profiling:
unified mass spectrometry and computational workflows for integrative proteomics–
metabolomics analysis. Mol. Omics 14, 307–319.
Brown, M.S., and Goldstein, J.L. (1997). The SREBP pathway: regulation of cholesterol
metabolism by proteolysis of a membrane-bound transcription factor. Cell 89, 331–340.
Brown, M.S., Ho, Y.K., and Goldstein, J.L. (1980). The cholesteryl ester cycle in
macrophage foam cells. Continual hydrolysis and re-esterification of cytoplasmic
cholesteryl esters. J. Biol. Chem. 255, 9344–9352.
Carlson, C.A., and Kim, K.-H. (1973). Regulation of Hepatic Acetyl Coenzyme A
Carboxylase by Phosphorylation and Dephosphorylation. J. Biol. Chem. 248, 378–380.
Chang, T.-Y., Reid, P.C., Sugii, S., Ohgami, N., Cruz, J.C., and Chang, C.C.Y. (2005).
Niemann-Pick Type C Disease and Intracellular Cholesterol Trafficking. Journal of
Biological Chemistry 280, 20917–20920.
Chang, T.Y., Li, B.L., Chang, C.C.Y., and Urano, Y. (2009). Acyl-coenzyme
Chapnick, D.A., and Liu, X. (2014). Leader cell positioning drives wound-directed
collective migration in TGFβ-stimulated epithelial sheets. Molecular Biology of the Cell
Chapnick, D.A., Bunker, E., and Liu, X. (2015). A biosensor for the activity of the
“sheddase” TACE (ADAM17) reveals novel and cell type-specific mechanisms of TACE
activation. Science Signaling 8.
Chapnick, D.A., Bunker, E., Liu, X., and Old, W.M. (2019). Temporal Metabolite, Ion,
and Enzyme Activity Profiling Using Fluorescence Microscopy and Genetically Encoded
Biosensors. Methods Mol. Biol. 1978, 343–353.
Cheng, D., Chang, C.C., Qu, X., and Chang, T.Y. (1995). Activation of acyl-coenzyme
A:cholesterol acyltransferase by cholesterol or by oxysterol in a cell-free system. J. Biol.
Chem. 270, 685–695.
Chong, J., Soufan, O., Li, C., Caraus, I., Li, S., Bourque, G., Wishart, D.S., and Xia, J.
(2018). MetaboAnalyst 4.0: Towards more transparent and integrative metabolomics
analysis. Nucleic Acids Research 46, W486–W494.
Clasquin, M.F., Melamud, E., and Rabinowitz, J.D. (2012). LC-MS data processing with
MAVEN: A metabolomic analysis and visualization engine. Current Protocols in
Dale, S., Wilson, W.A., Edelman, A.M., and Hardie, D.G. (1995). Similar substrate
recognition motifs for mammalian AMP-activated protein kinase, higher plant HMG-CoA
reductase kinase-A, yeast SNF1, and mammalian calmodulin-dependent protein kinase
I. FEBS Letters 361, 191–195.
Das, A., Brown, M.S., Anderson, D.D., Goldstein, J.L., and Radhakrishnan, A. (2014).
Three pools of plasma membrane cholesterol and their relation to cholesterol
homeostasis. Elife 3.
Dasari, A., Bartholomew, J.N., Volonte, D., and Galbiati, F. (2006). Oxidative Stress
Induces Premature Senescence by Stimulating Caveolin-1 Gene Transcription through
p38 Mitogen-Activated Protein Kinase/Sp1–Mediated Activation of Two GC-Rich
Promoter Elements. Cancer Res 66, 10805–10814.
Davignon, J. (2004). Beneficial cardiovascular pleiotropic effects of statins. Circulation
De Felice, B., Santillo, M., Serù, R., Damiano, S., Matrone, G., Wilson, R.R., and
Mondola, P. (2004). Modulation of 3-hydroxy-3-methylglutaryl-CoA reductase gene
expression by CuZn superoxide dismutase in human fibroblasts and HepG2 cells. Gene
Expr. 12, 29–38.
De Filippis, D., Esposito, G., Cirillo, C., Cipriano, M., De Winter, B.Y., Scuderi, C.,
Sarnelli, G., Cuomo, R., Steardo, L., De Man, J.G., et al. (2011). Cannabidiol reduces
intestinal inflammation through the control of neuroimmune axis. PLoS ONE 6, e28159.
Demais, V., Barthélémy, A., Perraut, M., Ungerer, N., Keime, C., Reibel, S., and
Pfrieger, F.W. (2016). Reversal of Pathologic Lipid Accumulation in NPC1-Deficient
Neurons by Drug-Promoted Release of LAMP1-Coated Lamellar Inclusions. J.
Neurosci. 36, 8012–8025.
Devinsky, O., Cross, J.H., Laux, L., Marsh, E., Miller, I., Nabbout, R., Scheffer, I.E.,
Thiele, E.A., and Wright, S. (2017). Trial of Cannabidiol for Drug-Resistant Seizures in
the Dravet Syndrome (Massachusetts Medical Society).
Dravet, C. (2011). The core Dravet syndrome phenotype. Epilepsia 52 Suppl 2, 3–9.
Esposito, G., Scuderi, C., Valenza, M., Togna, G.I., Latina, V., De Filippis, D., Cipriano,
M., Carratù, M.R., Iuvone, T., and Steardo, L. (2011). Cannabidiol reduces Aβ-induced
neuroinflammation and promotes hippocampal neurogenesis through PPARγ
involvement. PLoS ONE 6, e28668.
Fediuc, S., Gaidhu, M.P., and Ceddia, R.B. (2006). Regulation of AMP-activated protein
kinase and acetyl-CoA carboxylase phosphorylation by palmitate in skeletal muscle
cells. J. Lipid Res. 47, 412–420.
Ferreri, C. (2005). Life—As a Matter of Fat: The Emerging Science of Lipidomics.By Ole
G. Mouritsen. ChemBioChem 6, 1463–1464.
Gaston, T.E., Bebin, E.M., Cutter, G.R., Liu, Y., Szaflarski, J.P., and UAB CBD Program
(2017). Interactions between cannabidiol and commonly used antiepileptic drugs.
Epilepsia 58, 1586–1592.
Gelman, A., and Rubin, D.B. (1992). Inference from iterative simulation using multiple
sequences. Statistical Science 7, 457–472.
Geng, L., Zhou, W., Liu, B., Wang, X., and Chen, B. (2018). DHA induces apoptosis of
human malignant breast cancer tissues by the TLR-4/PPAR-α pathways. Oncol Lett 15,
George, K.S., and Wu, S. (2012). Lipid raft: A floating island of death or survival.
Toxicology and Applied Pharmacology 259, 311–319.
Ghovanloo, M.-R., Shuart, N.G., Mezeyova, J., Dean, R.A., Ruben, P.C., and
Goodchild, S.J. (2018). Inhibitory effects of cannabidiol on voltage-dependent sodium
currents. J. Biol. Chem. 293, 16546–16558.
Gianfrancesco, M.A., Paquot, N., Piette, J., and Legrand-Poels, S. (2018). Lipid bilayer
stress in obesity-linked inflammatory and metabolic disorders. Biochem. Pharmacol.
Giannini, E.G., Testa, R., and Savarino, V. (2005). Liver enzyme alteration: a guide for
clinicians. CMAJ 172, 367–379.
Gray, R.A., and Whalley, B.J. (2020). The proposed mechanism of action of CBD in
epilepsy. Epileptic Disorders : International Epilepsy Journal with Videotape 22, 10–15.
Gwinn, D.M., Shackelford, D.B., Egan, D.F., Mihaylova, M.M., Mery, A., Vasquez, D.S.,
Turk, B.E., and Shaw, R.J. (2008). AMPK Phosphorylation of Raptor Mediates a
Metabolic Checkpoint. Molecular Cell 30, 214–226.
Hafner, M., Mills, C.E., Subramanian, K., Chen, C., Chung, M., Boswell, S.A., Everley,
R.A., Liu, C., Walmsley, C.S., Juric, D., et al. (2019). Multiomics Profiling Establishes
the Polypharmacology of FDA-Approved CDK4/6 Inhibitors and the Potential for
Differential Clinical Activity. Cell Chemical Biology 26, 1067–1080.e8.
Hardie, D.G., Salt, I.P., Hawley, S.A., and Davies, S.P. (1999). AMP-activated protein
kinase: an ultrasensitive system for monitoring cellular energy charge. Biochem. J. 338 (
Pt 3), 717–722.
Hawley, S.A., Boudeau, J., Reid, J.L., Mustard, K.J., Udd, L., Mäkelä, T.P., Alessi, D.R.,
and Hardie, D.G. (2003). Complexes between the LKB1 tumor suppressor, STRAD
alpha/beta and MO25 alpha/beta are upstream kinases in the AMP-activated protein
kinase cascade. J. Biol. 2, 28.
Hawley, S.A., Pan, D.A., Mustard, K.J., Ross, L., Bain, J., Edelman, A.M., Frenguelli,
B.G., and Hardie, D.G. (2005). Calmodulin-dependent protein kinase kinase-β is an
alternative upstream kinase for AMP-activated protein kinase. Cell Metabolism 2, 9–19.
Hinz, B., and Ramer, R. (2019). Anti-tumour actions of cannabinoids. Br. J. Pharmacol.
Höglinger, D., Burgoyne, T., Sanchez-Heras, E., Hartwig, P., Colaco, A., Newton, J.,
Futter, C.E., Spiegel, S., Platt, F.M., and Eden, E.R. (2019). NPC1 regulates ER
contacts with endocytic organelles to mediate cholesterol egress. Nature
Communications 10, 4276.
Hsu, J.-L., Leu, W.-J., Hsu, L.-C., Liu, S.-P., Zhong, N.-S., and Guh, J.-H. (2018). Para-
Toluenesulfonamide Induces Anti-tumor Activity Through Akt-Dependent and -
Independent mTOR/p70S6K Pathway: Roles of Lipid Raft and Cholesterol Contents.
Front Pharmacol 9, 1223.
Ibeas Bih, C., Chen, T., Nunn, A.V.W., Bazelot, M., Dallas, M., and Whalley, B.J.
(2015). Molecular Targets of Cannabidiol in Neurological Disorders. Neurotherapeutics
Iffland, K., and Grotenhermen, F. (2017). An Update on Safety and Side Effects of
Cannabidiol: A Review of Clinical Data and Relevant Animal Studies. Cannabis and
Cannabinoid Research 2, 139–154.
Jow, H., Boys, R.J., and Wilkinson, D.J. (2014). Bayesian identification of protein
differential expression in multi-group isobaric labelled mass spectrometry data.
Statistical Applications in Genetics and Molecular Biology 13, 531–551.
Kenyon, J., Liu, W., and Dalgleish, A. (2018). Report of Objective Clinical Responses of
Cancer Patients to Pharmaceutical-grade Synthetic Cannabidiol. Anticancer Res. 38,
Kilsdonk, E.P., Yancey, P.G., Stoudt, G.W., Bangerter, F.W., Johnson, W.J., Phillips,
M.C., and Rothblat, G.H. (1995). Cellular cholesterol efflux mediated by cyclodextrins. J.
Biol. Chem. 270, 17250–17256.
Klein, U., Gimpl, G., and Fahrenholz, F. (1995). Alteration of the myometrial plasma
membrane cholesterol content with beta-cyclodextrin modulates the binding affinity of
the oxytocin receptor. Biochemistry 34, 13784–13793.
Krämer, A., Green, J., Pollard, J., and Tugendreich, S. (2014). Causal analysis
approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523–530.
Kuleshov, M.V., Jones, M.R., Rouillard, A.D., Fernandez, N.F., Duan, Q., Wang, Z.,
Koplev, S., Jenkins, S.L., Jagodnik, K.M., Lachmann, A., et al. (2016). Enrichr: a
comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids
Res. 44, W90-97.
Lange, Y., Ye, J., Rigney, M., and Steck, T.L. (1999). Regulation of endoplasmic
reticulum cholesterol by plasma membrane cholesterol. J. Lipid Res. 40, 2264–2270.
Lange, Y., Tabei, S.M.A., Ye, J., and Steck, T.L. (2013). Stability and Stoichiometry of
Bilayer Phospholipid–Cholesterol Complexes: Relationship to Cellular Sterol Distribution
and Homeostasis. Biochemistry 52, 6950–6959.
Lauckner, J.E., Jensen, J.B., Chen, H.-Y., Lu, H.-C., Hille, B., and Mackie, K. (2008).
GPR55 is a cannabinoid receptor that increases intracellular calcium and inhibits M
current. Proc. Natl. Acad. Sci. U.S.A. 105, 2699–2704.
Leek, J.T., Monsen, E., Dabney, A.R., and Storey, J.D. (2006). EDGE: extraction and
analysis of differential gene expression. Bioinformatics 22, 507–508.
Li, X., Wang, L., Zhou, X.E., Ke, J., de Waal, P.W., Gu, X., Tan, M.H.E., Wang, D., Wu,
D., Xu, H.E., et al. (2015). Structural basis of AMPK regulation by adenine nucleotides
and glycogen. Cell Research 25, 50–66.
Lindblom, G., and Orädd, G. (2009). Lipid lateral diffusion and membrane
heterogeneity. Biochim. Biophys. Acta 1788, 234–244.
Lingwood, D., and Simons, K. (2010). Lipid rafts as a membrane-organizing principle.
Science 327, 46–50.
Liou, G.I., Auchampach, J.A., Hillard, C.J., Zhu, G., Yousufzai, B., Mian, S., Khan, S.,
and Khalifa, Y. (2008). Mediation of cannabidiol anti-inflammation in the retina by
equilibrative nucleoside transporter and A2A adenosine receptor. Invest. Ophthalmol.
Vis. Sci. 49, 5526–5531.
Liu, M., Huang, W., Wu, D., and Priestley, J.V. (2006). TRPV1, but not P2X, requires
cholesterol for its function and membrane expression in rat nociceptors. European
Journal of Neuroscience 24, 1–6.
Lu, F., Liang, Q., Abi-Mosleh, L., Das, A., De Brabander, J.K., Goldstein, J.L., and
Brown, M.S. (2015). Identification of NPC1 as the target of U18666A, an inhibitor of
lysosomal cholesterol export and Ebola infection. ELife 4, e12177.
Lundbæk, J.A., Birn, P., Girshman, J., Hansen, A.J., and Andersen, O.S. (1996).
Membrane Stiffness and Channel Function. Biochemistry 35, 3825–3830.
Lundbaek, J.A., Birn, P., Tape, S.E., Toombes, G.E.S., Søgaard, R., Koeppe, R.E.,
Gruner, S.M., Hansen, A.J., and Andersen, O.S. (2005). Capsaicin regulates voltage-
dependent sodium channels by altering lipid bilayer elasticity. Mol. Pharmacol. 68, 680–
Massi, P., Solinas, M., Cinquina, V., and Parolaro, D. (2013). Cannabidiol as potential
anticancer drug. Br J Clin Pharmacol 75, 303–312.
McAllister, S.D., Soroceanu, L., and Desprez, P.-Y. (2015). The antitumor activity of
plant-derived non-psychoactive cannabinoids. J Neuroimmune Pharmacol 10, 255–267.
McFadden, J.W., and Corl, B.A. (2009). Activation of AMP-activated protein kinase
(AMPK) inhibits fatty acid synthesis in bovine mammary epithelial cells. Biochem.
Biophys. Res. Commun. 390, 388–393.
McPartland, J.M., Duncan, M., Di Marzo, V., and Pertwee, R.G. (2015). Are cannabidiol
and Δ(9) -tetrahydrocannabivarin negative modulators of the endocannabinoid system?
A systematic review. Br. J. Pharmacol. 172, 737–753.
Meyer-Arendt, K., Old, W.M., Houel, S., Renganathan, K., Eichelberger, B., Resing,
K.A., and Ahn, N.G. (2011). IsoformResolver: A Peptide-Centric Algorithm for Protein
Inference. Journal of Proteome Research 10, 3060–3075.
Miller, I., Scheffer, I.E., Gunning, B., Sanchez-Carpintero, R., Gil-Nagel, A., Perry, M.S.,
Saneto, R.P., Checketts, D., Dunayevich, E., Knappertz, V., et al. (2020). Dose-Ranging
Effect of Adjunctive Oral Cannabidiol vs Placebo on Convulsive Seizure Frequency in
Dravet Syndrome: A Randomized Clinical Trial. JAMA Neurol.
Mollinedo, F., and Gajate, C. (2015). Lipid rafts as major platforms for signaling
regulation in cancer. Adv Biol Regul 57, 130–146.
Munday, M.R. (2002). Regulation of mammalian acetyl-CoA carboxylase. Biochem Soc
Trans 30, 1059–1064.
Naito, T., Ercan, B., Krshnan, L., Triebl, A., Koh, D.H.Z., Wei, F.-Y., Tomizawa, K.,
Torta, F.T., Wenk, M.R., and Saheki, Y. (2019). Movement of accessible plasma
membrane cholesterol by the GRAMD1 lipid transfer protein complex. Elife 8.
Nemkov, T., Hansen, K.C., and D’Alessandro, A. (2017). A three-minute method for
high-throughput quantitative metabolomics and quantitative tracing experiments of
central carbon and nitrogen pathways. Rapid Communications in Mass Spectrometry
Nemkov, T., Reisz, J.A., Gehrke, S., Hansen, K.C., and D’Alessandro, A. (2019). High-
throughput metabolomics: Isocratic and gradient mass spectrometry-based methods. In
Methods in Molecular Biology, (Humana Press Inc.), pp. 13–26.
Norris, J.L., Farrow, M.A., Gutierrez, D.B., Palmer, L.D., Muszynski, N., Sherrod, S.D.,
Pino, J.C., Allen, J.L., Spraggins, J.M., Lubbock, A.L.R., et al. (2017). Integrated, High-
Throughput, Multiomics Platform Enables Data-Driven Construction of Cellular
Responses and Reveals Global Drug Mechanisms of Action. Journal of Proteome
Research 16, 1364–1375.
Ortegren, U., Karlsson, M., Blazic, N., Blomqvist, M., Nystrom, F.H., Gustavsson, J.,
Fredman, P., and Strålfors, P. (2004). Lipids and glycosphingolipids in caveolae and
surrounding plasma membrane of primary rat adipocytes. Eur. J. Biochem. 271, 2028–
Pan, H., Mukhopadhyay, P., Rajesh, M., Patel, V., Mukhopadhyay, B., Gao, B., Haskó,
G., and Pacher, P. (2009). Cannabidiol attenuates cisplatin-induced nephrotoxicity by
decreasing oxidative/nitrosative stress, inflammation, and cell death. J. Pharmacol. Exp.
Ther. 328, 708–714.
Pani, B., and Singh, B.B. (2009). Lipid rafts/caveolae as microdomains of calcium
signaling (Elsevier Ltd).
Pirmoradi, L., Seyfizadeh, N., Ghavami, S., Zeki, A.A., and Shojaei, S. (2019). Targeting
cholesterol metabolism in glioblastoma: a new therapeutic approach in cancer therapy.
J. Investig. Med. 67, 715–719.
Rajesh, M., Mukhopadhyay, P., Bátkai, S., Patel, V., Saito, K., Matsumoto, S.,
Kashiwaya, Y., Horváth, B., Mukhopadhyay, B., Becker, L., et al. (2010). Cannabidiol
attenuates cardiac dysfunction, oxidative stress, fibrosis, and inflammatory and cell
death signaling pathways in diabetic cardiomyopathy. J. Am. Coll. Cardiol. 56, 2115–
Reisz, J.A., Zheng, C., D’Alessandro, A., and Nemkov, T. (2019). Untargeted and Semi-
targeted Lipid Analysis of Biological Samples Using Mass Spectrometry-Based
Metabolomics. Methods Mol. Biol. 1978, 121–135.
Rimmerman, N., Juknat, A., Kozela, E., Levy, R., Bradshaw, H.B., and Vogel, Z. (2011).
The Non-Psychoactive Plant Cannabinoid, Cannabidiol Affects Cholesterol Metabolism-
Related Genes in Microglial Cells. Cellular and Molecular Neurobiology 31, 921–930.
Rimmerman, N., Ben-Hail, D., Porat, Z., Juknat, A., Kozela, E., Daniels, M.P., Connelly,
P.S., Leishman, E., Bradshaw, H.B., Shoshan-Barmatz, V., et al. (2013). Direct
modulation of the outer mitochondrial membrane channel, voltage-dependent anion
channel 1 (VDAC1) by cannabidiol: a novel mechanism for cannabinoid-induced cell
death. Cell Death & Disease 4, e949–e949.
Ross, H.R., Napier, I., and Connor, M. (2008). Inhibition of recombinant human T-type
calcium channels by Delta9-tetrahydrocannabinol and cannabidiol. J. Biol. Chem. 283,
Ryazanov, A.G., Shestakova, E.A., and Natapov, P.G. (1988). Phosphorylation of
elongation factor 2 by EF-2 kinase affects rate of translation. Nature 334, 170–173.
Ryberg, E., Larsson, N., Sjögren, S., Hjorth, S., Hermansson, N.O., Leonova, J.,
Elebring, T., Nilsson, K., Drmota, T., and Greasley, P.J. (2007). The orphan receptor
GPR55 is a novel cannabinoid receptor. British Journal of Pharmacology 152, 1092–
Schaffer, B.E., Levin, R.S., Hertz, N.T., Maures, T.J., Schoof, M.L., Hollstein, P.E.,
Benayoun, B.A., Banko, M.R., Shaw, R.J., Shokat, K.M., et al. (2015). Identification of
AMPK Phosphorylation Sites Reveals a Network of Proteins Involved in Cell Invasion
and Facilitates Large-Scale Substrate Prediction. Cell Metabolism 22, 907–921.
Schroeder, R., London, E., and Brown, D. (1994). Interactions between saturated acyl
chains confer detergent resistance on lipids and glycosylphosphatidylinositol (GPI)-
anchored proteins: GPI-anchored proteins in liposomes and cells show similar behavior.
Proc. Natl. Acad. Sci. U.S.A. 91, 12130–12134.
Serini, S., Trombino, S., Oliva, F., Piccioni, E., Monego, G., Resci, F., Boninsegna, A.,
Picci, N., Ranelletti, F.O., and Calviello, G. (2008). Docosahexaenoic acid induces
apoptosis in lung cancer cells by increasing MKP-1 and down-regulating p-ERK1/2 and
p-p38 expression. Apoptosis 13, 1172–1183.
Serra, M., and Saba, J.D. (2010). Sphingosine 1-phosphate lyase, a key regulator of
sphingosine 1-phosphate signaling and function. Adv. Enzyme Regul. 50, 349–362.
Shaw, R.J., Kosmatka, M., Bardeesy, N., Hurley, R.L., Witters, L.A., DePinho, R.A., and
Cantley, L.C. (2004). The tumor suppressor LKB1 kinase directly activates AMP-
activated kinase and regulates apoptosis in response to energy stress. Proc. Natl. Acad.
Sci. U.S.A. 101, 3329–3335.
Shin, S., Jing, K., Jeong, S., Kim, N., Song, K.-S., Heo, J.-Y., Park, J.-H., Seo, K.-S.,
Han, J., Park, J.-I., et al. (2013). The omega-3 polyunsaturated fatty acid DHA induces
simultaneous apoptosis and autophagy via mitochondrial ROS-mediated Akt-mTOR
signaling in prostate cancer cells expressing mutant p53. Biomed Res Int 2013, 568671.
Silvestri, C., Paris, D., Martella, A., Melck, D., Guadagnino, I., Cawthorne, M., Motta, A.,
and Di Marzo, V. (2015). Two non-psychoactive cannabinoids reduce intracellular lipid
levels and inhibit hepatosteatosis. Journal of Hepatology 62, 1382–1390.
Silvius, J.R. (2003). Role of cholesterol in lipid raft formation: lessons from lipid model
systems. Biochim. Biophys. Acta 1610, 174–183.
Smart, E.J., Ying, Y.S., Conrad, P.A., and Anderson, R.G. (1994). Caveolin moves from
caveolae to the Golgi apparatus in response to cholesterol oxidation. J. Cell Biol. 127,
Sokolov, A., and Radhakrishnan, A. (2010). Accessibility of Cholesterol in Endoplasmic
Reticulum Membranes and Activation of SREBP-2 Switch Abruptly at a Common
Cholesterol Threshold. J. Biol. Chem. 285, 29480–29490.
Staneva, G., Puff, N., Stanimirov, S., Tochev, T., Angelova, M.I., and Seigneuret, M.
(2018). The Alzheimer’s disease amyloid-β peptide affects the size-dynamics of raft-
mimicking Lo domains in GM1-containing lipid bilayers. Soft Matter 14, 9609–9618.
Sun, S.-N., Jia, W.-D., Chen, H., Ma, J.-L., Ge, Y.-S., Yu, J.-H., and Li, J.-S. (2013).
Docosahexaenoic acid (DHA) induces apoptosis in human hepatocellular carcinoma
cells. Int J Clin Exp Pathol 6, 281–289.
Supek, F., Bošnjak, M., Škunca, N., and Šmuc, T. (2011). REVIGO Summarizes and
Visualizes Long Lists of Gene Ontology Terms. PLOS ONE 6, e21800.
Thiele, E.A., Marsh, E.D., French, J.A., Mazurkiewicz-Beldzinska, M., Benbadis, S.R.,
Joshi, C., Lyons, P.D., Taylor, A., Roberts, C., Sommerville, K., et al. (2018).
Cannabidiol in patients with seizures associated with Lennox-Gastaut syndrome
(GWPCARE4): a randomised, double-blind, placebo-controlled phase 3 trial. Lancet
Trapnell, C., Williams, B.A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M.J.,
Salzberg, S.L., Wold, B.J., and Pachter, L. (2010). Transcript assembly and
quantification by RNA-Seq reveals unannotated transcripts and isoform switching during
cell differentiation. Nature Biotechnology 28, 511–515.
Venugopal, R., and Jaiswal, A.K. (1998). Nrf2 and Nrf1 in association with Jun proteins
regulate antioxidant response element-mediated expression and coordinated induction
of genes encoding detoxifying enzymes. Oncogene 17, 3145–3156.
Wang, G.L., and Semenza, G.L. (1993). General involvement of hypoxia-inducible
factor 1 in transcriptional response to hypoxia. Proc. Natl. Acad. Sci. U.S.A. 90, 4304–
Wang, Y., Mukhopadhyay, P., Cao, Z., Wang, H., Feng, D., Haskó, G., Mechoulam, R.,
Gao, B., and Pacher, P. (2017). Cannabidiol attenuates alcohol-induced liver steatosis,
metabolic dysregulation, inflammation and neutrophil-mediated injury. Scientific Reports
Wassall, S.R., Leng, X., Canner, S.W., Pennington, E.R., Kinnun, J.J., Cavazos, A.T.,
Dadoo, S., Johnson, D., Heberle, F.A., Katsaras, J., et al. (2018). Docosahexaenoic
acid regulates the formation of lipid rafts: A unified view from experiment and simulation.
Biochim Biophys Acta Biomembr 1860, 1985–1993.
Watkins, A.R. (2019). Cannabinoid interactions with ion channels and receptors.
Channels (Austin) 13, 162–167.
Whyte, L.S., Ryberg, E., Sims, N.A., Ridge, S.A., Mackie, K., Greasley, P.J., Ross,
R.A., and Rogers, M.J. (2009). The putative cannabinoid receptor GPR55 affects
osteoclast function in vitro and bone mass in vivo. Proc Natl Acad Sci U S A 106,
Widenmaier, S.B., Snyder, N.A., Nguyen, T.B., Arduini, A., Lee, G.Y., Arruda, A.P.,
Saksi, J., Bartelt, A., and Hotamisligil, G.S. (2017). NRF1 Is an ER Membrane Sensor
that Is Central to Cholesterol Homeostasis. Cell 171, 1094.e15–1109.e15.
Wiśniewski, J.R. (2016). Quantitative Evaluation of Filter Aided Sample Preparation
(FASP) and Multienzyme Digestion FASP Protocols. Anal. Chem. 88, 5438–5443.
Woods, A., Dickerson, K., Heath, R., Hong, S.-P., Momcilovic, M., Johnstone, S.R.,
Carlson, M., and Carling, D. (2005). Ca2+/calmodulin-dependent protein kinase kinase-
beta acts upstream of AMP-activated protein kinase in mammalian cells. Cell Metab. 2,
Wu, H.-Y., Goble, K., Mecha, M., Wang, C.-C., Huang, C.-H., Guaza, C., and Jan, T.-R.
(2012). Cannabidiol-induced apoptosis in murine microglial cells through lipid raft. Glia
Zhao, Y.-F., Wang, L., Lee, S., Sun, Q., Tuo, Y., Wang, Y., Pei, J., and Chen, C. (2010).
Cholesterol induces mitochondrial dysfunction and apoptosis in mouse pancreatic beta-
cell line MIN6 cells. Endocr 37, 76–82.