Marijuana Rx Corporation

HIV / AIDS Marijuana Treatment Clinical Research

HIV / AIDS Marijuana Treatment Clinical Research

Short-Term Effects of Cannabinoids in Patients with HIV-1 Infection: A Randomized, Placebo-Controlled Clinical Trial

Article (PDF Available) inAnnals of internal medicine 139(4):258-66 · September 2003with227 Reads

DOI: 10.7326/0003-4819-139-4-200308190-00008 · Source: PubMed
Cannabinoid use could potentially alter HIV RNA levels by two mechanisms: immune modulation or cannabinoid-protease inhibitor interactions (because both share cytochrome P-450 metabolic pathways). To determine the short-term effects of smoked marijuana on the viral load in HIV-infected patients. Randomized, placebo-controlled, 21-day intervention trial. The inpatient General Clinical Research Center at the San Francisco General Hospital, San Francisco, California. 67 patients with HIV-1 infection. Participants were randomly assigned to a 3.95%-tetrahydrocannabinol marijuana cigarette, a 2.5-mg dronabinol (delta-9-tetrahydrocannabinol) capsule, or a placebo capsule three times daily before meals. HIV RNA levels, CD4+ and CD8+ cell subsets, and pharmacokinetic analyses of the protease inhibitors. 62 study participants were eligible for the primary end point (marijuana group, 20 patients; dronabinol group, 22 patients; and placebo group, 20 patients). Baseline HIV RNA level was less than 50 copies/mL for 36 participants (58%), and the median CD4+ cell count was 340 x 109 cells/L. When adjusted for baseline variables, the estimated average effect versus placebo on change in log10 viral load from baseline to day 21 was -0.07 (95% CI, -0.30 to 0.13) for marijuana and -0.04 (CI, -0.20 to 0.14) for dronabinol. The adjusted average changes in viral load in marijuana and dronabinol relative to placebo were -15% (CI, -50% to 34%) and -8% (CI, -37% to 37%), respectively. Neither CD4+ nor CD8+ cell counts appeared to be adversely affected by the cannabinoids. Smoked and oral cannabinoids did not seem to be unsafe in people with HIV infection with respect to HIV RNA levels, CD4+ and CD8+ cell counts, or protease inhibitor levels over a 21-day treatment.

Figure 1. Flow of participants through the randomized trial.  


Table 1. Baseline Characteristics* 


Table 2. Changes in Viral Load Level by Group 


Table 3. Changes in CD4 and CD8 Cell Counts Relative to the Placebo Group* 


Full-text (PDF)

Available from: Starley Shade
Short-Term Effects of Cannabinoids in Patients with HIV-1 Infection
A Randomized, Placebo-Controlled Clinical Trial
Donald I. Abrams, MD; Joan F. Hilton, DSc; Roslyn J. Leiser, RN; Starley B. Shade, MPH; Tarek A. Elbeik, PhD; Francesca T. Aweeka, PharmD;
Neal L. Benowitz, MD; Barry M. Bredt, MA; Bradley Kosel, PharmD; Judith A. Aberg, MD; Steven G. Deeks, MD; Thomas F. Mitchell, MPH;
Kathleen Mulligan, PhD; Peter Bacchetti, PhD; Joseph M. McCune, MD, PhD; and Morris Schambelan, MD
Background: Cannabinoid use could potentially alter HIV RNA
levels by two mechanisms: immune modulation or cannabinoid–
protease inhibitor interactions (because both share cytochrome
P-450 metabolic pathways).
Objective: To determine the short-term effects of smoked mar-
ijuana on the viral load in HIV-infected patients.
Design: Randomized, placebo-controlled, 21-day intervention trial.
Setting: The inpatient General Clinical Research Center at the
San Francisco General Hospital, San Francisco, California.
Participants: 67 patients with HIV-1 infection.
Intervention: Participants were randomly assigned to a 3.95%-
tetrahydrocannabinol marijuana cigarette, a 2.5-mg dronabinol
(delta-9-tetrahydrocannabinol) capsule, or a placebo capsule three
times daily before meals.
Measurements: HIV RNA levels, CD4
and CD8
cell subsets,
and pharmacokinetic analyses of the protease inhibitors.
Results: 62 study participants were eligible for the primary end
point (marijuana group, 20 patients; dronabinol group, 22 pa-
tients; and placebo group, 20 patients). Baseline HIV RNA level
was less than 50 copies/mL for 36 participants (58%), and the
median CD4
cell count was 340 10
cells/L. When adjusted
for baseline variables, the estimated average effect versus placebo
on change in log
viral load from baseline to day 21 was #0.07
(95% CI, #0.30 to 0.13) for marijuana and #0.04 (CI, #0.20 to
0.14) for dronabinol. The adjusted average changes in viral load in
marijuana and dronabinol relative to placebo were #15% (CI,
#50% to 34%) and #8% (CI, #37% to 37%), respectively.
Neither CD4
nor CD8
cell counts appeared to be adversely
affected by the cannabinoids.
Conclusions: Smoked and oral cannabinoids did not seem to be
unsafe in people with HIV infection with respect to HIV RNA
levels, CD4
and CD8
cell counts, or protease inhibitor levels
over a 21-day treatment.
Ann Intern Med. 2003;139:258-266.
For author affiliations, see end of text.
arijuana has been smoked for medicinal purposes for
centuries (1). Introduced into western medicine in
1842, marijuana was used to treat various illnesses on the
basis of its purported analgesic, anticonvulsant, sedative,
hypnotic, and antispasmodic properties. With the passage
of the Marihuana Tax Act in 1937, use of marijuana as a
therapeutic agent in the United States waned until the
substance was removed from the U.S. Pharmacopoeia in
1942. The Controlled Substances Act of 1970 placed mar-
ijuana in the Schedule I category along with other sub-
stances deemed to have no medicinal value and high po-
tential for abuse.
In 1986, the U.S. Food and Drug Administration ap-
proved a synthetic, oral form of marijuana’s main psycho-
active component, delta-9-tetrahydrocannabinol (dronabi-
nol, Marinol, Roxane Laboratories, Columbus, Ohio), for
treating chemotherapy-induced nausea and vomiting (2–
5). A randomized, controlled trial demonstrated that dron-
abinol increased self-reported appetite but not weight in
patients with AIDS-related wasting syndrome; these find-
ings led to an expansion of the labeling indication for this
use in 1992 (6, 7). Before the advent of highly active an-
tiretroviral therapy in the 1990s, many patients infected
with HIV-1 experienced wasting as a preterminal manifes-
tation of the disease (8). Patients with AIDS-related wast-
ing syndrome often reported that they preferred smoked
marijuana to dronabinol because it was easier to titrate the
dose to achieve the desired effect; smoked marijuana deliv-
ers cannabinoids to the bloodstream much more rapidly
than dronabinol (9). By the mid-1990s, cannabis buyers’
clubs in the San Francisco Bay area were reportedly selling
marijuana to 11 000 patients with HIV infection (10–12).
With the increased availability of protease inhibitor–
containing antiretroviral regimens in the mid-1990s, the
incidence of AIDS-related wasting syndrome decreased
markedly, as did most of the other late-stage opportunistic
manifestations of advanced HIV disease (13–15). Protease
inhibitors, which can inhibit or stimulate the hepatic cyto-
chrome P-450 enzyme system, are subject to many signif-
icant drug–drug interactions with other agents used in
treating HIV infection and its complications (16, 17). The
potential for a drug– drug interaction between protease in-
hibitors and marijuana is worrisome since many HIV-
infected patients continue to smoke marijuana as an appe-
tite stimulant or to decrease nausea associated with their
antiretroviral therapy (18, 19). The likelihood of such an
interaction is supported by the facts that cannabinoids are
metabolized by some of the same cytochrome P-450 en-
zyme isoforms that metabolize the more widely prescribed
protease inhibitors and that tetrahydrocannabinol has been
shown to inhibit the metabolism of other drugs (20 –23).
Although few recent clinical trials have evaluated the
potential therapeutic effects of smoked marijuana, signifi-
cant progress has been made in understanding the pharma-
cology of cannabinoids in humans. Of the two cannabi-
noid receptors identified, CB1 (found mainly in cells of the
central nervous system) is thought to be responsible for the
neurologic and behavioral effects of marijuana (24, 25).
258 © 2003 American College of Physicians
The identication of a CB2 receptor, found predominantly
on B lymphocytes and natural killer cells, suggests that
cannabinoids may also affect the immune response. Some
studies suggest that marijuana can impair the immune sys-
tem through B-lymphocyte modulation, tumor necrosis
factor inhibition, or changes in the phenotype and func-
tion of circulating lymphocytes (26 29).
The hallmark of successful antiretroviral therapy is sus-
tained suppression of HIV RNA levels associated with in-
creasing CD4
cell counts (3032). Considering the
potential for both a protease inhibitorcannabinoid inter-
action and an effect of smoked marijuana on the immune
system, we designed a study to determine the safety or
toxicity prole of cannabinoids (smoked and oral) in per-
sons with HIV infection. We chose HIV RNA levels as our
primary outcome because an intervention that interacted
unfavorably with either the antiretroviral agent pharmaco-
kinetics or the immune system directly could cause a per-
turbation of viral suppression. We report the overall safety
results of this randomized, controlled inpatient clinical
Study Group
Study participants were recruited by referrals from lo-
cal physicians and advertisements in newspapers. Volun-
teers from across the country telephoned to determine
whether they might be eligible to participate. Participants
were required to be at least 18 years of age, have docu-
mented HIV infection, and be receiving a stable antiretro-
viral treatment regimen of either indinavir (Crixivan,
Merck & Co., Inc., North Wales, Pennsylvania) or nel
navir (Viracept, Agouron Pharmaceuticals, Inc., La Jolla,
California) for at least 8 weeks before enrollment. When
enrolled, participants who had been taking the recently
recommended dose of nelnavir, 1250 mg twice daily,
were switched to 750 mg three times daily for consistency
of our pharmacokinetic evaluations (33). No additional
protease inhibitors were allowed for the duration of the
study. Participants were also required to have a stable viral
load, dened as less than a threefold (0.5 log
) change in
HIV RNA level for the 16 weeks before enrollment. All
participants were required to have previous experience
smoking marijuana (dened as six or more times) to ensure
that they knew how to inhale and what neuropsychiatric
effects to expect. The institutional review board of the
University of California, San Francisco, approved the
study, and signed, informed consent was obtained from
each participant before enrollment.
Exclusion criteria included any active opportunistic in-
fection or malignant condition requiring acute treatment,
unintentional loss of 10% or more of body weight during
the previous 6 months, current substance dependence as-
certained by completion of a condential drug screening
form and an alcohol screening form, methadone mainte-
nance, use of tobacco or cannabinoids (smoked or oral)
within 30 days of enrollment, history of serious pulmonary
disease, pregnancy, or stage II or higher AIDS dementia
complex. Laboratory exclusion criteria were hematocrit less
than 0.25 and elevation of hepatic aminotransferase levels
to greater than ve times the upper limit of normal. Ther-
apeutic exclusions were concurrent use within the past 8
weeks of anabolic hormones, prednisone, interleukin-2, or
other agents known to alter immune system function.
Study Medications
The National Institute on Drug Abuse provided pre-
rolled marijuana cigarettes, weighing on average 0.9 g and
containing 3.95% delta-9-tetrahydrocannabinol. These
cigarettes were kept in a locked and alarmed freezer until
they were dispensed to a locked freezer in the General
Clinical Research Center at the San Francisco General
Hospital, where the inpatient study was conducted. The
frozen marijuana cigarettes required rehydration overnight
in a humidier. Participants randomly assigned to the
smoked marijuana group were housed in a room with a fan
ventilating to the outside. To maximize standardization of
inhaled doses, research staff monitored participants while
they followed the uniform puff procedure outlined by Fol-
tin and colleagues (34). Research staff weighed the mari-
juana cigarettes immediately before and after they were
administered to participants and returned all leftover ma-
terial to the pharmacy. Study participants smoked up to
three complete marijuana cigarettes daily, as tolerated, 1
hour before meals. Study participants were randomly as-
signed in a double-blind fashion to the oral regimens,
which were given on the same schedule as the smoked
marijuana. Research staff observed participants taking all
Because the same systems metabolize cannabinoids and
protease inhibitors, cannabinoids might alter viral loads in
HIV-infected patients taking protease inhibitors.
In this randomized trial, 62 HIV-infected patients taking
indinavir or nelfinavir received a marijuana cigarette, dron-
abinol capsule, or placebo capsule three times daily for 21
days. Half of the patients in all three groups had undetect-
able viral loads during the study, and average changes in
viral load with marijuana and dronabinol, relative to pla-
cebo, were small.
The findings of no large harmful effects on viral loads with
either smoked or oral cannabinoids need to be confirmed
in larger and longer trials.
–The Editors
ArticleCannabinoids in Patients with HIV Infection 19 August 2003 Annals of Internal Medicine Volume 139 Number 4 259
Research Design and Procedures
Study clinicians admitted study participants to the
General Clinical Research Center for a 4-day lead-in pe-
riod to obtain baseline variables. A urine sample obtained
on the day of admission (day 4) had to be negative for
tetrahydrocannabinol. The second phase of the trial was a
21-day intervention period beginning with random assign-
ment of treatments on day 0. Patients were stratied by
protease inhibitor (indinavir or nelnavir) and then allo-
cated with equal probability in blocks of 12 to the study
agents (marijuana, dronabinol, and placebo). The statisti-
cian generated the random allocation sequences, and the
pharmacists maintained the sequences in a secure location
and distributed the assignments to the study coordinator
on day 0.
Study participants were not permitted to have visitors
or to leave the General Clinical Research Center unless
accompanied by research personnel during the 25-day
study. All clinical laboratory tests and study procedures
were obtained or performed in the center. Patients were
weighed on the same calibrated scale each morning while
wearing a hospital gown.
Baseline blood specimens were collected on days 4
and 0 to examine within-participant variation in HIV
RNA level in the absence of experimental therapies.
Follow-up specimens were obtained on days 2, 5, 8, 11,
14, 17, 19, and 21. Samples were stored at 70 °C and
batch-tested for HIV RNA at the end of the trial by using
branched DNA (bDNA) technology (VERSANT HIV-1
RNA 3.0 Assay, Bayer Diagnostics, Emeryville, California)
with a lower detection limit of 50 copies/mL.
Baseline samples for CD4
and CD8
cell counts
were collected on days 4 and 0, and follow-up specimens
were drawn on days 7, 14, and 21. Assays were performed
in the San Francisco General Hospital Clinical Laboratory.
Complete blood counts with differential were performed
by using an automated hematology analyzer (Bayer Tech-
nicon H3 system, Bayer Corp., Tarrytown, New York)
according to the manufacturers directions. The CD4
and CD8
cell counts were measured by using Multi
TEST CD3/CD8/CD45/CD4 with Trucount tubes (BD
Biosciences, San Jose, California) according to the manu-
facturers directions. Data acquisition and analysis were
performed by using a FACSCalibur (BD Biosciences) ow
cytometer and MultiSET software (BD Biosciences).
Pharmacokinetic methods are described elsewhere (35).
Statistical Analysis
This randomized trial was designed to compare the
marijuana and dronabinol groups with the placebo group
with respect to mean changes in log
HIV RNA levels
between days 0 and 21. We planned the sample size for
two one-sided Bonferroni-adjusted 0.05-level t-tests of the
null hypothesis of no difference against the alternative that
the cannabinoid effect is larger than 0.3 log
each with 80% power. This design, which assumed an SD
of 0.3 log
copies/mL for within-participant changes, re
quired 21 participants per group. To allow for potential
dropouts, we enrolled two additional patients per group.
The between-group difference of 0.3 log
copies/mL rep
resents a doubling of the viral load on the natural scale and
a clinically signicant and potentially unsafe effect of can-
nabinoid on HIV RNA levels (30). Changes less than 0.3
copies/mL are considered to be within the natural
range of variability of log
HIV RNA measurements (36,
To evaluate the success of the randomization proce-
dures, we examined the distributions by group of several
baseline variables, including CD4
and CD8
cell counts
and HIV RNA levels on day 0 and protease inhibitor used.
When a participants viral load level was undetectable, a
value of 49 copies/mL was assumed. HIV RNA levels were
transformed to the log
scale, and each participants
change in viral load level on day 21 relative to day 0 was
calculated. We summarized the raw changes by group by
using means, 95% CIs of differences between mean
changes, and P values. We used multiple regression to
model the cannabinoid effects while controlling for the
effects of baseline covariates, including age (#40 years, 40
to 49 years, and $49 years), race or ethnicity (white, Af-
rican American, Latino, or other), protease inhibitor, viral
load detectability on day 0, small or large RNA change
during the lead-in period (!0.5 versus $0.5 log
mL), and baseline log
and log
counts. Similarly, we modeled log
HIV RNA levels at
day 0 and all eight follow-up time points, using a random
intercept repeated-measures model. This model allowed
baseline covariates to modify either the intercept or the
slope and included a quadratic time trend for patients with
large RNA changes during the lead-in period. This sub-
group showed marked benet from participation in the
clinical trial during the lead-in period and early part of the
follow-up period; their RNA levels were typical of all study
participants. The simpler model compared HIV RNA lev-
els at the start and end of the trial (two levels per partici-
pant), whereas the repeated-measures model used nine lev-
els per participant to estimate the changes from day 0 to
day 21; therefore, the latter cannabinoid effect estimates
were less inuenced by measurement error at any one time
point. Because we were concerned about violations of
model assumptions of normality and homoscedasticity, all
CIs and P values reported were calculated by using the
bias-corrected, accelerated bootstrap method with partici-
pant-level resampling and 2000 bootstrap iterations (38).
These are valid even when the assumptions are violated.
Finally, each model was examined for the effects of inu-
ential observations, identied through the algorithm of Le-
saffre and Verbeke (39).
The cannabinoid groups also were compared with the
placebo group with respect to changes in CD4
cell counts, adjusted for the covariates above and
for baseline HIV RNA level. The model of CD4
Article Cannabinoids in Patients with HIV Infection
260 19 August 2003 Annals of Internal Medicine Volume 139 Number 4
counts was additionally adjusted for baseline CD8
and vice versa. We added 10 to the cell counts to reduce
the inuence of very small values and then transformed to
the log
scale to ensure model validity. These models es
timate multiplicative effects on geometric means, which we
described as percentage effects by converting the effect on
the original log scale with the formula (10
1) %
100%. For example, an effect of 0.05 is a 12% greater
increase in cell count for a cannabinoid than a placebo
participant with the same initial count, regardless of
whether it was 0.005 or 0.5 % 10
cells/L. We used medi
ans and ranges to describe within-group changes in body
weight over the study period and MannWhitney tests to
compare the cannabinoid and placebo groups. All P values
reported are two-sided.
To investigate the effect of imputing a single xed
value of 49 copies/mL for undetectable viral loads, we used
the SAS Lifereg procedure (SAS Institute, Inc., Cary,
North Carolina) to instead treat undetectable viral loads as
left-censored at the detection limit. Although this method
is usually used for survival time analysis, we obtained the
needed models by using viral load as the time variable and
specifying a log-normal distribution.
Role of the Funding Source
The funding source reviewed and funded the protocol
and provided the marijuana cigarettes for the trial.
Characteristics of Patients
A total of 603 individuals volunteered for the study,
but most did not meet the eligibility criteria (Figure 1). Of
the 69 study participants admitted to the inpatient study
unit, 67 were randomly assigned between May 1998 and
May 2000. Thirty-seven patients were receiving nelnavir-
containing regimens and 30 patients were receiving indi-
navir-containing regimens. Of these, 3 and 2 patients, re-
spectively, left the study before the pharmacokinetic
analysis on day 14. The remaining 62 study participants
completed the 21-day inpatient intervention phase and
were eligible for all end points (marijuana group, 20 pa-
tients; dronabinol group, 22 patients; and placebo group,
20 patients).
Most patients were men (89%) older than 40 years of
age (68%), and half were of nonwhite ethnicity (Table 1).
More patients in the marijuana and dronabinol groups
Figure 1. Flow of participants through the randomized trial.
THC & tetrahydrocannabinol.
ArticleCannabinoids in Patients with HIV Infection 19 August 2003 Annals of Internal Medicine Volume 139 Number 4 261
than in the placebo group had previous AIDS diagnoses
and detectable HIV RNA than in the placebo group. Over-
all, 58% of the participants had undetectable HIV RNA
levels (#50 copies/mL); only 5 patients had HIV RNA
levels greater than 10 000 copies/mL, 4 of whom were
receiving nelnavir-containing regimens. Baseline CD4
and CD8
cell counts were similar in all groups.
During the 4-day lead-in phase, no participants HIV
RNA level increased by 0.5 log
copies/mL (3.2-fold).
However, HIV RNA levels decreased by at least this
amount in 5 patients (marijuana group, 3 patients; dron-
abinol group, 2 patients; placebo group, 0 patients): 1 of
28 patients receiving indinavir, 1 of 13 patients receiving
nelnavir three times daily, and 3 of 21 patients originally
receiving nelnavir twice daily. Changing the nelnavir
regimen from two to three doses per day seemed to have a
large effect on HIV RNA levels. However, since large de-
creases in HIV RNA occurred in participants receiving all
three regimens, they also might be due to the fact that
therapy was directly observed.
Change in HIV RNA Levels
HIV RNA was undetectable at days 0 and 21 in 50%
to 55% of patients in each group (Table 2). Although the
median change in each group was 0, the mean changes
were decreases in both cannabinoid groups: marijuana
group, 0.14 log
copies/mL (95% CI, 0.42 to 0.03
copies/mL), and dronabinol group, 0.18 log
ies/mL (CI, 0.51 to 0.04 log
copies/mL). These nd
ings were due mainly to ve study participants with 0.5
copies/mL or greater decreases in viral load during
follow-up. The mean change among patients receiving pla-
cebo, 0.06 log
copies/mL (CI, 0.03 to 0.24 log
ies/mL), was an increase, and no patient experienced a
large decrease during follow-up. The unadjusted mean
change in the marijuana group was 0.19 log
(CI, 0.48 to 0.01 log
copies/mL) lower than in the
placebo group, and the corresponding mean difference be-
tween the dronabinol and placebo groups was 0.24 log
copies/mL (CI, 0.55 to 0.06 log
copies/mL). After
we controlled for the large change in HIV RNA level dur-
ing the lead-in period (!0.5 vs. $0.5 log
decrease) and
other covariates previously mentioned, the mean marijua-
naplacebo difference was 0.06 log
copies/mL (CI,
0.26 to 0.13 log
copies/mL) and the mean dronabi
nolplacebo difference was 0.07 log
copies/mL (CI,
0.24 to 0.06 log
copies/mL). Models treating unde
tectable viral loads as left-censored produced slightly higher
upper condence bounds of 0.23 for the marijuanapla-
cebo difference and 0.09 for the dronabinolplacebo dif-
The repeated-measures models of nine measurements
per study participant seemed to t adequately with only
linear terms for treatment effects over time, since quadratic
terms did not approach statistical signicance. A quadratic
term was needed only for the ve patients with large
change in HIV RNA level during the lead-in period. Be-
fore adjustment, the cannabinoids seemed to reduce viral
load, whereas after adjustment they seemed to have little
effect on this outcome. In particular, on the basis of the
adjusted model, both upper condence bounds for the
treatment effects (marijuana group, 0.13 [34%]; dronabi-
nol group, 0.14 [37%]) excluded cannabinoid-associated
Table 1. Baseline Characteristics*
Characteristic Marijuana Group
$ 20)
Dronabinol Group
$ 22)
Placebo Group
$ 20)
All Groups (
$ 62)
Median age (range), y 41.5 (33–54) 43 (34–52) 44.5 (26–80) 43 (26–80)
Sex, n (%)
Men 17 (85) 19 (86) 19 (95) 55 (89)
Women 2 (10) 1 (5) 0 (0) 3 (5)
Transgender (male-to-female) 1 (5) 2 (9) 1 (5) 4 (6)
Race or ethnicity, n (%)
White 13 (65) 9 (41) 9 (45) 31 (50)
African American 3 (15) 6 (27) 3 (15) 12 (19)
Latino or Latina 1 (5) 4 (18) 5 (25) 10 (16)
Other 3 (15) 3 (14) 3 (15) 9 (15)
Median body mass index (range), kg/m
25.6 (21.9–53.3) 25.0 (14.8–38.2) 25.4 (18.7–33.0) 25.5 (14.8–53.3)
Use of protease inhibitor, n (%)
Nelfinavir 11 (55) 12 (55) 11 (55) 34 (55)
Indinavir 9 (45) 10 (45) 9 (45) 28 (45)
Previous opportunistic infection or malignant condition, n (%) 12 (60) 12 (55) 6 (30) 30 (48)
Median HIV RNA level (range), log
3.5 (2.0–4.5) 3.5 (1.7–4.3) 3.7 (1.8–4.6) 3.6 (1.7–4.6)
Undetectable HIV RNA levels, n (%) 12 (60) 11 (50) 13 (45) 36 (58)
Median CD4
cell count (range), %10
0.345 (0.026–0.9) 0.315 (0.052–0.771) 0.378 (0.007–0.906) 0.34 (0.007–0.906)
cell count # 200 % 10
cells/L, n (%)
5 (25) 5 (24) 5 (28) 15 (24)
Median CD8
cell count (range), %10
0.736 (0.433–1.987) 0.91 (0.223–2.23) 0.708 (0.3–1.987) 0.757 (0.223–2.23)
* Among patients with baseline viral load levels $ 50 copies/mL.
Three patients had missing data for CD4
and CD8
cell counts on day 0: dronabinol group, 1 patient, and placebo group, 2 patients.
Article Cannabinoids in Patients with HIV Infection
262 19 August 2003 Annals of Internal Medicine Volume 139 Number 4
increases in viral load of 0.3 log
copies/mL (100%), our
a priori threshold for concern.
Change in CD4
and CD8
Cell Subsets
Figure 2 shows the median changes in absolute num-
bers of CD4
and CD8
cells over the 21-day experimen
tal intervention. Compared with patients receiving
placebo, the unadjusted mean increases in CD4
counts were greater for patients receiving cannabinoids
than for patients receiving placebo (marijuana group, 20%
[CI, 7% to 55%]; dronabinol group, 17% [CI, 5% to
45%]) (Table 3). The adjusted two-point model and the
repeated-measures model showed similar ndings.
Over the 21-day follow-up period, increases in CD8
cell counts were on average 20% (CI, 7% to 38%) greater
for patients receiving marijuana than for patients receiving
placebo and marginally greater (10% [CI, 5% to 29%])
for patients receiving dronabinol than for those receiving
placebo. In the adjusted repeated-measures model, the can-
nabinoid effects were similar (lower condence bounds:
marijuana group, 4%; dronabinol group, 3%). An anal-
ysis of expanded immune system phenotypes and functions
revealed few statistically signicant effects (40).
The detailed results of the effects of the cannabinoids
on the pharmacokinetics of the protease inhibitors have
been described elsewhere (35, 41). No clinically signicant
alterations of nelnavir or indinavir levels were noted.
Change in Weight
Although safety was the primary end point of this trial,
study participants underwent many evaluations to assess
the effect of cannabinoids on appetite, caloric intake,
weight, and body composition. Over the 21-day study pe-
riod, the placebo recipients gained a median of 1.1 kg
(range, 1.4 to 5.2 kg). The participants in the marijuana
and dronabinol groups gained signicantly more weight, a
median of 3.0 kg (range, 0.75 to 8.6 kg; P & 0.021) and
3.2 kg (range, 1.4 to 7.6 kg; P & 0.004), respectively.
Dual-energy x-ray absorptiometry demonstrated that most
of the weight gained in all groups was fat mass (42).
Figure 2. Changes in CD4
and CD8
cell counts by group
$ 62).
Top. Median change in CD4
cell counts over the 21-day study period.
Bottom. Median change in CD8
cell counts over the 21-day study
Table 2. Changes in Viral Load Level by Group
Variable Marijuana Group (
$ 20) Dronabinol Group (
$ 22) Placebo Group (
$ 20)
Change between day 0 and day 21 (2 time points), n (%)
Increase $ 0.5 log
1 (5) 0 (0) 1 (5)
Increase 0.5 log
4 (20) 2 (9) 5 (25)
Decrease # 0.5 log
2 (10) 7 (32) 3 (15)
Decrease 0.5 log
3 (15) 2 (9) 0 (0)
No change 10 (50) 11 (50) 11 (55)
Unadjusted mean difference in viral load from placebo group
(95% CI), log
0.19 (0.48 to 0.01)* 0.24 (0.55 to 0.06)† —
Adjusted mean difference in viral load from placebo group
(95% CI)‡, log
0.06 (0.26 to 0.13)§ 0.07 (0.24 to 0.06)§ —
Average change in viral load at day 21 (repeated measures:
9 time points), log
Adjusted mean difference in viral load from placebo group
(95% CI)! 0.07 (0.30 to 0.13)§ 0.04 (0.20 to 0.14)§ —
* P & 0.07.
P # 0.001.
Multivariable models were adjusted for the following covariates: age, race, protease inhibitor, viral load detectability at day 0, small or large viral load change during the
lead-in period, baseline log
cell counts, and log
cell counts. Three patients who were missing data on baseline CD4
and CD8
cell counts were excluded
from multivariate models. The models yielded results similar to those of the models that included all independent variables and led to the same conclusions.
§ P $ 0.2.
! In addition to the covariates listed, this model controlled for a quadratic time effect among patients with large viral load change during the lead-in period.
ArticleCannabinoids in Patients with HIV Infection 19 August 2003 Annals of Internal Medicine Volume 139 Number 4 263
This study provides evidence that short-term use of
cannabinoids, either oral or smoked, does not substantially
elevate viral load in individuals with HIV infection who
are receiving stable antiretroviral regimens containing nel
navir or indinavir. Upper condence bounds for all esti-
mated effects of cannabinoids on HIV RNA level from all
analyses were no greater than an increase of 0.23 log
copies/mL compared with placebo. Because this study was
randomized and analyses were controlled for all known
potential confounders, it is very unlikely that chance im-
balance on any known or unknown covariate masked a
harmful effect of cannabinoids. Study participants in all
groups may have been expected to benet from the equiv-
alent of directly observed antiretroviral therapy, as well as
decreased stress and, for some, improved nutrition over the
25-day inpatient stay.
Neither CD4
nor CD8
cell counts seemed to be
adversely affected by the cannabinoids during the study;
lower condence bounds on estimated cannabinoid effects
typically exceeded 0, indicating benet rather than harm.
Increases in CD8
cell counts in the marijuana group seen
in our study differ from ndings reported in earlier studies
conducted in participants without HIV infection (29). The
clinical signicance and mechanism accounting for these
changes are unclear.
The pharmacokinetic component of this study did not
demonstrate clinically signicant interactions with canna-
binoids that would warrant dose adjustments of protease
inhibitors in the context of smoked marijuana or dronabi-
nol use (35). However, given the great variability of the
pharmacokinetics of protease inhibitors, the long-term sig-
nicance of the short-term concentration decreases ob-
served is not known.
Although the primary objective of this study was to
assess the safety of cannabinoids in patients with HIV in-
fection treated with protease inhibitorcontaining antiret-
roviral regimens, a secondary aim was to obtain some in-
formation on activity, particularly about appetite stimulation
and weight gain. Whereas previous studies of dronabinol
have demonstrated signicantly increased appetite and
only a trend toward weight gain, this trial shows increased
weight in both cannabinoid groups compared with the pla-
cebo group. However, the weight gained by the cannabi-
noid recipients was not in the desired lean body mass but
in fat.
Our conclusions are limited by the short duration of
this study. Also, few women participated, so our results
may apply mainly to men. The results of this study, which
evaluated government-supplied marijuana of known po-
tency and content, cannot be extrapolated to the potential
effects of marijuana available on the street. In addition, the
lack of a blinded control group for the smoked marijuana
arm could bias the interpretation of some of our results,
such as the weight changes; however, it is difcult to at-
tribute effects on HIV RNA level and CD4
and CD8
cell counts to any such potential bias. We chose not to
include a smoked placebo group because we thought it
would be impossible to blind marijuana in study partici-
pants with previous marijuana experience. Of interest,
most of the patients receiving dronabinol (17 of 22) could
identify their blinded treatment correctly, whereas the pa-
tients in the placebo group had more difculty (9 of 20).
This suggests that placebo-controlled studies of the efcacy
of smoked marijuana could be considered in the future.
The Institute of Medicine reviewed accumulated data on
the safety and effectiveness of marijuana as medicine in a re-
cent comprehensive report (43). The discussion of medicinal
marijuana is a polarizing one that is confounded by emotion
and politics, usually unsupported by data. Our short-duration
clinical trial suggests acceptable safety in a vulnerable im-
mune-compromised patient population. Further studies in-
vestigating the therapeutic potential of marijuana and other
cannabinoids in patients with HIV infection and other pop-
ulations are ongoing and should provide additional safety in-
formation over longer exposure periods (44).
From the University of California, San Francisco, and Gladstone Insti-
tute of Virology and Immunology, San Francisco, California; and the
Washington University School of Medicine, St. Louis, Missouri.
Acknowledgments: The authors thank the research nursing and dietary
staff at the San Francisco General Hospital General Clinical Research
Center for the professionalism and compassion with which they con-
ducted the trial. They also appreciate the efforts of the San Francisco
General Hospital inpatient research pharmacy staff and are deeply in-
Table 3. Changes in CD4
and CD8
Cell Counts Relative to
the Placebo Group*
Variable Marijuana
$ 20)
$ 21)
Relative change in CD4
cell count between
day 0 and day 21 (2 time points)
Unadjusted estimated effect, % 20 (7 to 55) 17 (5 to 45)
P value #0.001 #0.001
Adjusted estimated effect, %† 13 (1 to 28) 12 (2 to 28)
P value 0.06 0.09
Relative change in CD4
cell count at day
21 (repeated measures: 4 time points)
Adjusted estimated effect, %† 16 (2 to 33) 14 (1 to 32)
P value 0.025 0.064
Relative change in CD8
cell count between
day 0 and day 21 (2 time points)
Unadjusted estimated effect, % 20 (7 to 38) 10 (5 to 29)
P value 0.002 $0.2
Adjusted estimated effect, %† 16 (2 to 36) 8 (5 to 27)
Relative change in CD8
cell count at day
21 (repeated measures: 4 time points)
Adjusted estimated effect, %† 20 (4 to 42) 10 (3 to 32)
P value 0.016 0.15
* The placebo group included 18 participants. All values in parentheses are 95%
Multivariable models included the following covariates: age; race; protease in-
hibitor; viral load detectability at day 0; small or large viral load change during the
lead-in period; baseline log
HIV RNA level; and baseline log
and log
cell counts for log
and log
cell models, respectively. The
models yielded results similar to those of the models that included all independent
variables and led to the same conclusions.
Article Cannabinoids in Patients with HIV Infection
264 19 August 2003 Annals of Internal Medicine Volume 139 Number 4
debted to the committed study participants. The authors also thank
Bayer Diagnostics for providing the VERSANT HIV-1 RNA 3.0 Assays
and disposables and Roxane Laboratories for the dronabinol and placebo
capsules. Finally, the authors thank Dr. Jag H. Khalsa of the Center on
AIDS and Other Medical Consequences of Drug Abuse at the National
Institute on Drug Abuse for his thoughtful guidance and constant sup-
port, and Rick Doblin, PhD, for his inspiration and persistence.
Grant Support: By National Institutes of Health grants 1RO1 DA/
MH11607, 5-MO1-RR00083, and P30-MH59037.
Potential Financial Conflicts of Interest: Consultancies: N.L. Benowitz
(Alexza Molecular Delivery Corp.); Honoraria: D.I. Abrams (Solvay
Pharmaceuticals), T.A. Elbeik (Bayer Diagnostics), F.T. Aweeka (Merck
& Co.); Stock ownership or options (other than mutual funds): N.L. Be-
nowitz (Alexza Molecular Delivery Corp.); Grants received: T.A. Elbeik
(Bayer Diagnostics, Cellestics Ltd., Roche Molecular Systems), F.T.
Aweeka (Agouron Pharmaceuticals).
Requests for Single Reprints: Donald I. Abrams, MD, University of
California, San Francisco, Positive Health Program, Ward 84, 995 Po-
trero Avenue, San Francisco, CA 94110.
Current author addresses and author contributions are available at www
1. Abel EL. Marijuana: The First Twelve Thousand Years. New York: Plenum
Pr; 1980.
2. Sallan SE, Zinberg NE, Frei E 3rd. Antiemetic effect of delta-9-tetrahydro-
cannabinol in patients receiving cancer chemotherapy. N Engl J Med. 1975;293:
795-7. [PMID: 1099449]
3. Chang AE, Shiling DJ, Stillman RC, Goldberg NH, Seipp CA, Barofsky I,
et al. Delata-9-tetrahydrocannabinol as an antiemetic in cancer patients receiving
high-dose methotrexate. A prospective, randomized evaluation. Ann Intern Med.
1979;91:819-24. [PMID: 293141]
4. Frytak S, Moertel CG, O’Fallon JR, Rubin J, Creagan ET, O’Connell MJ, et
al. Delta-9-tetrahydrocannabinol as an antiemetic for patients receiving cancer
chemotherapy. A comparison with prochlorperazine and a placebo. Ann Intern
Med. 1979;91:825-30. [PMID: 517882]
5. Chang AE, Shiling DJ, Stillman RC, Goldberg NH, Seipp CA, Barofsky I,
et al. A prospective evaluation of delta-9-tetrahydrocannabinol as an antiemetic in
patients receiving adriamycin and cytoxan chemotherapy. Cancer. 1981;47:1746-
51. [PMID: 6261926]
6. Gorter R, Seefried M, Volberding P. Dronabinol effects on weight in patients
with HIV infection [Letter]. AIDS. 1992;6:127. [PMID: 1311935]
7. Beal JE, Olson R, Laubenstein L, Morales JO, Bellman P, Yangco B, et al.
Dronabinol as a treatment for anorexia associated with weight loss in patients
with AIDS. J Pain Symptom Manage. 1995;10:89-97. [PMID: 7730690]
8. Kotler DP, Tierney AR, Wang J, Pierson RN Jr. Magnitude of body-cell-
mass depletion and the timing of death from wasting in AIDS. Am J Clin Nutr.
1989;50:444-7. [PMID: 2773823]
9. Agurell S, Halldin M, Lindgren JE, Ohlsson A, Widman M, Gillespie H, et
al. Pharmacokinetics and metabolism of delta 1-tetrahydrocannabinol and other
cannabinoids with emphasis on man. Pharmacol Rev. 1986;38:21-43. [PMID:
10. Abrams DI, Child CC, Mitchell T. Marijuana, the AIDS wasting syndrome,
and the U.S. government [Letter]. N Engl J Med. 1995;333:670-1.
11. Abrams DI. Medical marijuana: tribulations and trials. J Psychoactive Drugs.
1998;30:163-9. [PMID: 9692378]
12. Abrams DI. Potential interventions for HIV/AIDS wasting: an overview. J
Acquir Immune Dec Syndr. 2000;25 Suppl 1:S74-80. [PMID: 11126431]
13. Palella FJ Jr, Delaney KM, Moorman AC, Loveless MO, Fuhrer J, Satten
GA, et al. Declining morbidity and mortality among patients with advanced
human immunodeciency virus infection. HIV Outpatient Study Investigators.
N Engl J Med. 1998;338:853-60. [PMID: 9516219]
14. Mocroft A, Vella S, Benfield TL, Chiesi A, Miller V, Gargalianos P, et al.
Changing patterns of mortality across Europe in patients infected with HIV-1.
EuroSIDA Study Group. Lancet. 1998;352:1725-30. [PMID: 9848347]
15. Ledergerber B, Egger M, Erard V, Weber R, Hirschel B, Furrer H, et al.
AIDS-related opportunistic illnesses occurring after initiation of potent antiretro-
viral therapy: the Swiss HIV Cohort Study. JAMA. 1999;282:2220-6. [PMID:
16. Acosta EP, Kakuda TN, Brundage RC, Anderson PL, Fletcher CV. Phar-
macodynamics of human immunodeciency virus type 1 protease inhibitors.
Clin Infect Dis. 2000;30 Suppl 2:S151-9. [PMID: 10860900]
17. Piscitelli SC, Gallicano KD. Interactions among drugs for HIV and oppor-
tunistic infections. N Engl J Med. 2001;344:984-96. [PMID: 11274626]
18. Child CC, Mitchell TF, Abrams DI. Patterns of therapeutic marijuana use in
two community-based cannabis buyerscooperatives [Abstract]. 12th World
AIDS Conference, Geneva, 28 June3 July 1998. Abstract no. 60569:1105.
19. Braitstein P, Kendall TR, Chan K, Montaner JSG, O’Shaughnessy MV,
Hogg RS. Mary-Jane and her patients: sociodemographic and clinical character-
istics of HIV! individuals using medicinal marijuana and antiretrovirals in Brit-
ish Columbia, Canada [Abstract]. XIIIth International AIDS Conference,
Durban, South Africa, 914 July 2000. Page 326, abstract no. ThPeB5055.
20. Yamamoto I, Watanabe K, Narimatsu S, Yoshimura H. Recent advances in
the metabolism of cannabinoids. Int J Biochem Cell Biol. 1995;27:741-6.
[PMID: 7584607]
21. Watanabe K, Matsunaga T, Yamamoto I, Funae Y, Yoshimura H. Involve-
ment of CYP2C in the metabolism of cannabinoids by human hepatic micro-
somes from an old woman. Biol Pharm Bull. 1995;18:1138-41. [PMID:
22. Benowitz NL, Jones RT. Effects of delta-9-tetrahydrocannabinol on drug
distribution and metabolism. Antipyrine, pentobarbital, and ethanol. Clin Phar-
macol Ther. 1977;22:259-68. [PMID: 891094]
23. Benowitz NL, Nguyen TL, Jones RT, Herning RI, Bachman J. Metabolic
and psychophysiologic studies of cannabidiol-hexobarbital interaction. Clin Phar-
macol Ther. 1980;28:115-20. [PMID: 7389248]
24. Pertwee RG. Pharmacology of cannabinoid CB1 and CB2 receptors. Phar-
macol Ther. 1997;74:129-80. [PMID: 9336020]
25. Felder CC, Glass M. Cannabinoid receptors and their endogenous agonists.
Annu Rev Pharmacol Toxicol. 1998;38:179-200. [PMID: 9597153]
26. Nahas GG, Suciu-Foca N, Armand JP, Morishima A. Inhibition of cellular
mediated immunity in marihuana smokers. Science. 1974;183:419-20. [PMID:
27. Friedman H, Klein TW, Newton C, Daaka Y. Marijuana, receptors and
immunomodulation. Adv Exp Med Biol. 1995;373:103-13. [PMID: 7668140]
28. Hollister LE. Marijuana and immunity. J Psychoactive Drugs. 1992;24:159-
64. [PMID: 1506999]
29. Wallace JM, Tashkin DP, Oishi JS, Barbers RG. Peripheral blood lympho-
cyte subpopulations and mitogen responsiveness in tobacco and marijuana smok-
ers. J Psychoactive Drugs. 1988;20:9-14. [PMID: 3392635]
30. U.S. Public Health Service. Guidelines for the use of antiretroviral agents
in HIV-infected adults and adolescents. Accessed at
/guidelines/default_db2.asp?id&50 on 17 February 2003.
31. Yeni PG, Hammer SM, Carpenter CC, Cooper DA, Fischl MA, Gatell JM,
et al. Antiretroviral treatment for adult HIV infection in 2002: updated recom-
mendations of the International AIDS Society-USA Panel. JAMA. 2002;288:
222-35. [PMID: 12095387]
32. Dybul M, Fauci AS, Bartlett JG, Kaplan JE, Pau AK. Guidelines for using
antiretroviral agents among HIV-infected adults and adolescents. Ann Intern
Med. 2002;137:381-433. [PMID: 12617573]
33. Johnson M, Petersen A, Winslade J, Clendennin N. Comparison of BID
and TID dosing of VIRACEPT
(nelnavir, NFV) in combination with stavu
dine (d4T) and lamivudine (3TC) [Abstract]. Proceedings of the 5th Conference
on Retroviruses and Opportunistic Infections, Chicago, Illinois, 15 February
1998. Page 148, abstract no. 373.
34. Foltin RW, Fischman MW, Byrne MF. Effects of smoked marijuana on
food intake and body weight of humans living in a residential laboratory. Appe-
tite. 1988;11:1-14. [PMID: 3228283]
ArticleCannabinoids in Patients with HIV Infection 19 August 2003 Annals of Internal Medicine Volume 139 Number 4 265
35. Kosel BW, Aweeka FT, Benowitz NL, Shade SB, Hilton JF, Lizak PS, et al.
The effects of cannabinoids on the pharmacokinetics of indinavir and nelnavir.
AIDS. 2002;16:543-50. [PMID: 11872997]
36. Hughes MD, Johnson VA, Hirsch MS, Bremer JW, Elbeik T, Erice A, et
al. Monitoring plasma HIV-1 RNA levels in addition to CD4
count improves assessment of antiretroviral therapeutic response. ACTG 241
Protocol Virology Substudy Team. Ann Intern Med. 1997;126:929-38. [PMID:
37. Paxton WB, Coombs RW, McElrath MJ, Keefer MC, Hughes J, Sinangil
F, et al. Longitudinal analysis of quantitative virologic measures in human im-
munodeciency virus-infected subjects with $ or & 400 CD4 lymphocytes:
implications for applying measurements to individual patients. National Institute
of Allergy and Infectious Diseases AIDS Vaccine Evaluation Group. J Infect Dis.
1997;175:247-54. [PMID: 9203644]
38. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. London: Chap-
man and Hall; 1993:178-88, 214-8, 398-403.
39. Lesaffre E, Verbeke G. Local influence in linear mixed models. Biometrics.
1998;54:570-82. [PMID: 9629645]
40. Bredt BM, Higuera-Alhino D, Shade SB, Hebert SJ, McCune JM, Abrams
DI. Short-term effects of cannabinoids on immune phenotype and function in
HIV-1-infected patients. J Clin Pharmacol. 2002;42:82S-89S. [PMID:
41. Kane B. Medical marijuana: the continuing story. Ann Intern Med. 2001;
134:1159-62. [PMID: 11412075]
42. Mulligan K, Abrams DI, Leiser RJ, Schambelan M. Body composition
changes in HIV-infected men consuming self-selected diets during a placebo-
controlled inpatient study of cannabinoids [Abstract]. Proceedings of the 8th
Conference on Retroviruses and Opportunistic Infections, Chicago, Illinois, 48
February 2001. Page 238, abstract no. 647.
43. Joy JE, Watson SJ, Benson JA, eds. Marijuana and Medicine: Assessing the
Science Base. Washington, DC: National Academy Pr; 1999.
44. University of California Center for Medicinal Cannabis Research. Accessed
at on 17 February 2003.
Article Cannabinoids in Patients with HIV Infection
266 19 August 2003 Annals of Internal Medicine Volume 139 Number 4
Current Author Addresses: Drs. Abrams and Deeks: University of Cal-
ifornia, San Francisco, Positive Health Program, Ward 84, 995 Potrero
Avenue, San Francisco, CA 94110.
Dr. Hilton: 500 Parnassus Avenue, San Francisco, CA 94143-0560.
Ms. Leiser: University of California, San Francisco, 143 Stillings Avenue,
San Francisco, CA 94131.
Mr. Shade and Mr. Mitchell: University of California, San Francisco,
3180 18th Street, Suite 201, San Francisco, CA 94110.
Drs. Elbeik and Benowitz and Mr. Bredt: University of California, San
Francisco, 1001 Potrero Avenue, San Francisco, CA 94110.
Dr. Aweeka: University of California, San Francisco, 521 Parnassus Av-
enue, San Francisco, CA 94143-0622.
Dr. Kosel: University of Washington, 325 9th Avenue, Seattle, WA
Dr. Aberg: Washington University, 4511 Forest Park, St. Louis, MO
Dr. Mulligan: San Francisco General Hospital, 1001 Potrero Avenue,
Building 30, R3501D, San Francisco, CA 94110.
Dr. Bacchetti: University of California, San Francisco, 500 Parnassus
Avenue, Room 423A West, San Francisco, CA 94143-0560.
Dr. McCune: Gladstone Institute, PO Box 419100, San Francisco, CA
Dr. Schambelan: San Francisco General Hospital, Building 5, 5B6, San
Francisco, CA 94110.
Author Contributions: Conception and design: D.I. Abrams, J.F.
Hilton, R.J. Leiser, S.B. Shade, T.A. Elbeik, F.T. Aweeka, N.L. Beno-
witz, T.F. Mitchell, K. Mulligan, J.M. McCune, M. Schambelan.
Analysis and interpretation of the data: D.I. Abrams, J.F. Hilton, S.B.
Shade, T.A. Elbeik, F.T. Aweeka, N.L. Benowitz, B.M. Bredt, B. Kosel,
K. Mulligan, P. Bacchetti, J.M. McCune, M. Schambelan.
Drafting of the article: D.I. Abrams, J.F. Hilton, T.A. Elbeik, S.G.
Deeks, T.F. Mitchell, J.M. McCune.
Critical revision of the article for important intellectual content: D.I.
Abrams, J.F. Hilton, R.J. Leiser, S.B. Shade, T.A. Elbeik, F.T. Aweeka,
N.L. Benowitz, B.M. Bredt, B. Kosel, J.A. Aberg, S.G. Deeks, T.F.
Mitchell, K. Mulligan, P. Bacchetti, J.M. McCune, M. Schambelan.
Final approval of the article: D.I. Abrams, S.B. Shade, F.T. Aweeka,
N.L. Benowitz, B.M. Bredt, J.A. Aberg, S.G. Deeks, T.F. Mitchell, K.
Mulligan, P. Bacchetti, J.M. McCune, M. Schambelan.
Provision of study materials or patients: D.I. Abrams, R.J. Leiser, T.A.
Elbeik, J.A. Aberg, S.G. Deeks.
Statistical expertise: J.F. Hilton, S.B. Shade, P. Bacchetti.
Obtaining of funding: D.I. Abrams, J.F. Hilton, T.A. Elbeik, T.F.
Administrative, technical, or logistic support: R.J. Leiser, S.B. Shade,
T.A. Elbeik, B.M. Bredt, J.A. Aberg, S.G. Deeks, T.F. Mitchell, M.
Collection and assembly of data: R.J. Leiser, T.A. Elbeik, F.T. Aweeka,
B.M. Bredt, B. Kosel, S.G. Deeks, K. Mulligan, M. Schambelan

Cannabinoid Receptor 2-Mediated Attenuation of CXCR4-Tropic HIV Infection in Primary CD4+ T Cells

Yuntao Wu, Editor


Agents that activate cannabinoid receptor pathways have been tested as treatments for cachexia, nausea or neuropathic pain in HIV-1/AIDS patients. The cannabinoid receptors (CB1R and CB2R) and the HIV-1 co-receptors, CCR5 and CXCR4, all signal via Gαi-coupled pathways. We hypothesized that drugs targeting cannabinoid receptors modulate chemokine co-receptor function and regulate HIV-1 infectivity. We found that agonism of CB2R, but not CB1R, reduced infection in primary CD4+ T cells following cell-free and cell-to-cell transmission of CXCR4-tropic virus. As this change in viral permissiveness was most pronounced in unstimulated T cells, we investigated the effect of CB2R agonism on to CXCR4-induced signaling following binding of chemokine or virus to the co-receptor. We found that CB2R agonism decreased CXCR4-activation mediated G-protein activity and MAPK phosphorylation. Furthermore, CB2R agonism altered the cytoskeletal architecture of resting CD4+ T cells by decreasing F-actin levels. Our findings suggest that CB2R activation in CD4+ T cells can inhibit actin reorganization and impair productive infection following cell-free or cell-associated viral acquisition of CXCR4-tropic HIV-1 in resting cells. Therefore, the clinical use of CB2R agonists in the treatment of AIDS symptoms may also exert beneficial adjunctive antiviral effects against CXCR4-tropic viruses in late stages of HIV-1 infection.


Cannabinoid agonists are currently under investigation for the treatment of AIDS-associated cachexia, nausea, and neuropathic pain . One such drug, dronabinol (Δ9-THC; Marinol®), has won Food and Drug Administration (FDA) approval for treatment of HIV-associated anorexia . Additionally, the prescription of smoked or ingested cannabis (marijuana) for treatment of AIDS-related symptoms has been approved in 14 states . Despite the use of cannabinoids by HIV/AIDS patients, few studies have investigated the impact of such drugs in regard to viral pathogenesis or immune regulation. Early studies conducted in the pre-HAART era suggested a positive correlation between development of opportunistic infection, progression to AIDS, and marijuana use . Yet recent analysis of HIV/AIDS patients enrolled a randomized, placebo-controlled clinical trial designed to study the outcome of cannabinoid administration have indicated that cannabinoid use does not result in significant immunosuppression . Indeed, both smoked marijuana and dronabinol were reported to increase total CD4+ T cell number  and naïve T cell number  over a 21-day period. A decrease in viral load was also observed in these patients . Similarly, in SIV infected rhesus macaques, Δ9-THC exposure reduced viral load and CD4+ T cell depletion, significantly increasing animal survival over an 11 month period . Despite these findings, the mechanisms by which cannabinoid drugs can influence viral replication or pathogenicity remain unknown.

Cannabinoid agonists activate the CB1R and CB2R cannabinoid receptors. Like the HIV chemokine co-receptors CXCR4 and CCR5, CB1R and CB2R are members of the Gαi-coupled family A GPCRs . CB2R is highly expressed on all CD4+ T cells , whereas CB1 expression is found in activated, memory subsets . CB1 and CB2 have been classified as immunosuppressive receptors on CD4+ T cells , although antagonism of CB1R and CB2R does not enhance immune activation and knock-out mice do not exhibit differences in T cell frequency or increases in autoimmune pathogenesis . The mechanism(s) by which cannabinoid agonists can modulate CD4+ T cell function remain unclear. Activation of CB2R has been shown to inhibit inflammatory cytokine production in CD4+ T cells , which may account for the decrease in autoimmune pathogenesis observed in therapeutic trials of cannabinoid agonists in animal models of multiple sclerosis . CB2R may also function as a chemotactic modulator, as CB2R activation inhibits CXCR4-induced chemotaxis in transformed lymphocytes . CB2R has further been shown to regulate lymphocyte egress from the bone marrow in a role previously attributed largely to CXCR4 . These findings suggest that CB2R may play a role in regulating chemokine receptor signaling, specifically the activity of CXCR4. Such cross-talk between CB2R and CXCR4 may have implications for AIDS patients who take cannabinoid-derived agents for therapeutic purposes.

Although coreceptor signaling is not essential for HIV-1 infection, several recent studies have suggested that chemokine receptor signaling enhances infection of resting CD4+ T cells . These cells express CXCR4, but not CCR5, whose expression is restricted to a small subset of memory CD4+ T cells . In patients, the emergence of CXCR4-tropic virus usually occurs after years of infection and correlates with more rapid progression to AIDS . Viral conversion to CXCR4-tropism increases the number of targets available to the virus . Additionally, as HIV-1 can establish latency in resting T cells , a switch to CXCR4-tropism could enhance the establishment of a latent pools of virus within lymphoid tissues. The increased number of new targets may explain the rapid decline in CD4+ T cell numbers and increased viral load in late-stage AIDS patients with CXCR4-tropic virus . The late-stage patients who frequently harbor CXCR4-tropic virus are also the most likely to benefit from cannabinoid drug use. It is therefore relevant to study the potential for cannabinoid signaling to modulate CXCR4 activity and alter the course of HIV infection, Interactions between GPCRs like CB2R and CXCR4 can cause cross-desensitization, allosteric modulation, dimerization, changes in receptor localization, and alteration of physiological function among GPCR pairings . Given that direct antagonism of chemokine receptor function can block viral infection , it is possible that allosteric modulation of CXCR4 through a GPCR partner may also reduce HIV-1 permissiveness. Indeed, oligomerization of the chemokine co-receptors including CXCR4 using conformationally specific monoclonal antibodies can inhibit HIV-1 entry into target cells . These experiments demonstrate signaling-independent modulation of coreceptor function. Allosteric agents that disrupt HIV-1 infection by modulating chemokine receptor signaling have not yet been identified. Should CB2R-induced signaling alter CXCR4 co-receptor function, this would represent first known example of a signaling-dependent GPCR interaction leading to viral inhibition.

To test this hypothesis, we examined the effect of cannabinoid receptor activation on HIV viral transmission and productive infection in CD4+ T cells using a GFP-expressing, CXCR4-tropic HIV-1 variant. We found that activation of CB2R on CD4+ T cells significantly inhibited viral infection in a CB2R -selective and dose-dependent manner. Viral inhibition was more pronounced in resting cells that were activated after infection. We investigated signaling in these cells and found that CB2R agonism significantly decreased SDF-1α-induced CXCR4 activation. Furthermore, CB2R agonism altered HIV-induced cytoskeletal rearrangement, associated with productive HIV infection in resting cells. We found that CB2R was not sufficient to block viral transfer or fusion, but did significantly diminish productive viral infection. We conclude that CB2R is a novel modulator of CXCR4-tropic HIV infection in CD4+ T cells.

Materials and Methods

Cell culture reagents

Anonymous blood donations were obtained from New York Blood Center. Cells were cultured as previously reported . The purified monoclonal antibodies (mAbs) 〈CD2 (OKT3) and 〈CD28 (28.2) were purchased from eBioscience. Phytohemagglutinin (PHA) and carboxyfluorescein succinimidyl ester (CFSE) was purchased from Sigma. Recombinant human IL-2 and CXCR4 antagonist AMD3100 were obtained through the AIDS Research and Reference Reagent Program (NIH). Fluorophore-conjugated CXCR4 (12GS), CCR5 (T21/8), CD25 (BC96), and CD45RO were purchased from Biolegend. The CB2 agonists JWH-133, JWH-150, Ser160, 2-arachidonoylglycerol, and anandamide and the CB2 antagonist AM630 were purchased from Tocris.

Cell purification and sorting

Total CD4+ T cells were isolated from healthy HIV/hepatitis B virus-seronegative donors as described previously . For FACS-sorting, CD4+ T cells were labeled with Live/Dead for viability (Invitrogen), stained for CD45RO, and sorted with a FACS Aria (BD Biosciences).

Viral constructs

HIV NL-GI and Gag-iCherry are NL4-3 based CXCR4-tropic HIV-1 molecular clones that have been described previously . NL-GI expresses green fluorescent protein (GFP) in place of the viral early gene nef, and nef expression is maintained by insertion of an internal ribosome entry site (IRES) . Gag-iCherry carries the GFP variant Cherry inserted internally into Gag between the MA and CA domains. For CCR5-tropic virus, a variant of NL-GI expressing the Env gene from the molecular clone JRFL  was used. Virus was produced in HEK293T cells and p24 concentration was calculated by ELISA prior to use.

Cell-Free Infection Assay

CD4+ T cells were thawed, resuspended in RPMI medium containing 20 U/ml recombinant IL-2 and stimulated with 1 µg/ml PHA (Sigma) overnight. Cells were cultured for four days and reseeded into 96 well flat-bottom plates (Costar) at a density of 105 cells/well prior to treatment and infection. Treated cells were incubated with antagonist or vehicle (DMSO) for one hour at 37°C followed by incubation with agonist or vehicle (DMSO or 0.1% ethanol) for another three hours. Following this treatment, triplicate cultures were infected with 10 ng/well HIV NL-GI. To assess HIV infection, fluorescence was assessed at day 4 post-treatment. Harvested cells were stained for viability and fixed with 4% paraformaldehyde, prior to acquisition on a FACSCalibur (BD Biosciences) and analysis with FloJo software (TreeStar).

Cell-Associated Infection Assay

HIV-expressing Jurkat donor cells were generated by transfection using HIV Gag-iCherry, as described previously . To generate infected CD4+ T cell donors, PHA activated CD4+ T cells were spinoculated with either HIV NL-GI or for 90 minutes at 1200×g. After 24–48 hours, approximately 10–30% of donor cells were infected. Donor cells were labeled with 10 µM CellTracker blue (CMF2HC) fluorescent dye (Invitrogen) and then co-cultured with unstimulated and antagonist and/or agonist-treated target cells in a 1∶1 ratio for 3 hours at 37°C, as described previously . Virus transfer was terminated by washing with PBS and treatment with 0.05% trypsin-EDTA (Invitrogen) for 5 minutes. Cells were stained for viability and for CD45RO prior to fixation and acquisition on a LSRII (BD Biosciences). Sorted GFPCherryCD45RO+ or CD45RO targets were seeded into 96-well plates coated with 2.5 µg/ml anti-CD3 in RPMI media containing 1 µg/ml anti-CD28 and 20 U/ml rIL-2 for activation. After 4 days, cells were harvested, fixed, and analyzed for fluorescence.

Quantitation of Viral Membrane Fusion

Cell-free viral fusion was measured using a method described previously . BlaM-VPR was a gift from Michael Miller (Merck Research Laboratories). Viral infections were done with 20 ng of virus at a concentration of 200 ng/ml for 2 hrs at 37°C. Cells were analyzed on an LSRII flow cytometer (Becton Dickinson).

Signaling Studies

For GTPγS binding, JWH-133 treated CD4+ T cells were permeabilized with CHAPS and incubated with increasing concentrations of SDF-1α. Incorporation of radiolabelled GTPγS was assayed as described . For kinase phosphorylation and western blotting, primary CD4+ T cells were treated with antagonist and/or agonist in serum-free RPMI for a total of 4 hours prior to treatment with SDF-1α (PeproTech) or NL4.3 HIV at various times indicated. Levels of phosphorylated MAP kinase and total MAP kinase were determined as described .

Data Analysis

Data was analyzed with GraphPad Prism using a paired t-test. Mean and standard error indicate variance between multiple donors, with n as indicated.


CB2-specific agonists inhibit HIV-1 infection in primary CD4+ T cells

As activation of the cannabinoid receptor CB2R has been shown to modify chemokine receptor activity  and cannabinoid use in HIV-1 infected individuals is associated with reductions in viral load , we hypothesized that CB2R agonism may alter the course of HIV-1 infection. Using a GFP-expressing variant of the CXCR4-tropic lab isolate NL4.3 (NL-GI), in which GFP is expressed in place of the viral gene nef , we assayed HIV-1 infectivity in primary CD4+ T cells pre-treated with CB2R agonist. Purified, blood-derived CD4+ T cells were stimulated with low-dose mitogen for four days prior to a three-hour cannabinoid treatment. To directly test the capacity for CB2R activation to inhibit viral infection, we pretreated activated CD4+ T cells with a potent and selective CB2R agonist that is approximately 200-fold selective for CB2R over CB1R (Ki = 3.4 nM)  (JWH-133) prior to HIV-1 exposure. Treated cells were then washed and exposed to virus in suspension (10 ng/2×105 cells) for four days. After this time, the frequency of GFP+ infected cells was measured and the frequency of inhibition, as compared to control (DMSO treated) infection, was calculated.

When cells were treated with 100 nM of JWH-133 prior to viral exposure, we observed an approximately 40% reduction in HIV-1 infected cells after four days (Figure 1A–B). This inhibition was significantly reduced when CB2R activity was selectively blocked using the antagonist AM630, indicating that the antiviral activity of JWH-133 is indeed CB2R-specific (Figure 1A–B). JWH-133-mediated blockade of HIV-1 infectivity was dose-dependent, with an EC50 of 7.59±0.1 nM and a plateau of efficacy at approximately 50% inhibition (Figure 1B). Further, the action of this drug was CXCR4-specific, as JWH-133 treatment was not sufficient to inhibit infection with an isogenic virus that carried a CCR5-tropic Env from molecular clone JRFL (Figure 1C). Therefore, CB2 activation reduces CXCR4-tropic HIV-1 infection in primary CD4+ T cells in a dose-dependent and receptor-specific fashion.

Figure 1

Cannabinoid inhibition of CXCR4-tropic HIV-1 infection is dose-dependent and CB2-selective.

To further confirm this novel role for CB2R, we tested other CB2R-selective agonists (JWH-015 and Ser016) to see if treatment was sufficient to reduce viral infectivity. Both of the highly specific CB2R agonists we tested proved to be antiviral at concentrations of 1 µM; these were JWH-015 (35.9±11.96% inhibition) and Ser016 (30.7±13.72% inhibition) (mean±SEM, n = 8 donors) (Figure 1D). Treatment with 1 µM of the pan-cannabinoid agonist Hu210 also significantly reduced HIV-1 infection, although to a lesser extent than with CB2R-selective agonists (23.9±7.45%, mean±SEM, n = 8 donors) (Figure 1D). The reduction in infection efficiency observed with Hu210 was CB2R-specific and was not observed when cells were pretreated with the CB2R-selective antagonist AM630 (Figure 1D). Consistent with these findings, 1 µM pretreatment with the CB1R selective agonist arachidonyl-2′-chloroethylamide (ACEA) did not significantly reduce viral infection (4.76±3.13%, mean±SEM, n = 5 donors) (Figure 1D). These results indicate that CB2R-selective agonists, but not CB1R-selective agonists, can inhibit HIV-1 permissiveness.

Given that the pan-cannabinoid agonist Hu210 possessed antiviral activity, we predicted that naturally occurring endogenous ligands of the cannabinoid receptors, the endocannabinoids 2-arachidonoylglycerol (2-AG) and anandamide (AEA) could reduce infectivity via CB2R activation. We found that pretreatment with either 2-AG or AEA significantly inhibited HIV-1 infection in a dose-dependent and CB2-specific manner (Figure 1E–F). AEA, which is reported to have higher affinity for CB2R than 2-AG (Ki = 371 nM versus 1400 nM, respectively) , proved to be a more potent inhibitor of virus, with an EC50 of 8.59±0.09 nM, as compared to 1.82±0.25 µM for 2-AG. The antiviral activity of both of these endocannabinoid agents was abrogated by treatment with 1 µM of the CB2R-selective antagonist AM630. These data demonstrate that, like synthetic cannabinoid agonists, the endocannabinoids can activate CB2R to inhibit HIV-1 infectivity in CD4+ T cells.

CB2 activation does not alter CXCR4 expression or T cell activation

We next sought to determine the mechanism of HIV-1 inhibition via CB2R. Previous reports have indicated that cannabinoid treatment of immune cells can lead to changes in cell surface chemokine receptor expression , loss of proliferative capacity , and reduction in effector function . We examined whether CB2R activation in our infection model led to either a reduction in co-receptor expression or in host fitness, rendering the cells incapable of harboring productive viral infection. To test this possibility, we treated CD4+ T cells with the CB2 agonist JWH-133 with a concentration of drug sufficient to inhibit viral infection (100 nM). We found that this treatment did not lead to significant reduction of CXCR4 cell surface expression (Figure 2A–B) or total CXCR4 protein expression (data not shown).

Figure 2

CB2 agonism at concentrations sufficient to inhibit HIV-1 infection does not significantly alter cell-surface CXCR4 expression or inhibit cellular proliferation in primary CD4+ T cells.

Likewise, pretreatment with up to 1 µM of JWH-133 prior to TCR-mediated activation did not reduce T cell activation, as measured by an increase in CD25 expression, or proliferation, as indicated by CFSE dilution after stimulation with anti-CD3 and anti-CD28 antibodies (Figure 2C–E). Concentrations of JWH-133 ranging from 1 µM and below did not lead to a change in cell viability as compared to DMSO treated controls (data not shown). Higher concentrations of JWH-133 or Hu210 (10 µM and above) did lead to apoptosis and cell death, consistent with published observations . Our findings indicate that low doses of the CB2R agonist JWH-133 are sufficient to inhibit viral infection without significant disruption of CD4+ T cell co-receptor expression, CD4 expression, CD25 upregulation, or proliferation. Further, CB2R agonism did not alter the ability of the CD4+ T cell to support viral infection, as JWH-133 treated cells are readily infected by the CCR5-tropic JRFL virus (Figure 1C). Taken together, these data suggest that CB2R activation is altering a pathway specifically required for CXCR4-tropic infection.

The antiviral activity of CB2 agonist JWH-133 is enhanced in unstimulated CD4+ T cells

Despite intense study, the importance of GPCR-mediated signaling during CXCR4-tropic HIV-1 infection remains unclear. Recently, several lines of study have demonstrated a role for chemokine receptor signaling in resting CD4+ T cells . Activation of Gαi-coupled chemokine receptors enhances infectivity in resting CD4+ T cells that are stimulated with gp120 or chemokine agonist after exposure to HIV-1 . Conversely, inhibition of GPCR function by pertussis toxin inhibits viral infection . This has been described as a model for latent infection in resting cells . We tested to see if the CB2 agonist JWH-133 differentially inhibited HIV-1 infection in resting versus activated cells.

Cells were activated with mitogen four days before (activated) (Figure 3A) or four days after (resting) (Figure 3B) treatment with 100 nM JWH-133 or 1 µM pertussis toxin and exposure to virus. We found that JWH-133 and pertussis toxin partially inhibited viral replication in both activated and resting cultures (Figure 3). Consistent with previous reports , inhibition by pertussis toxin was augmented in resting cultures as compared to activated cultures (59.38±29.5% versus 41.16±20.6%, respectively, *p<0.05) (mean±SEM, n = 4). Similarly, we found that the efficacy of JWH-133–mediated viral inhibition was significantly increased in resting, as compared to activated, cultures (63.34±31.7% versus 36.31±18.2%, respectively, *p<0.05 (mean±SEM, n = 4) (Figure 3C). These data are consistent with the premise that alteration of GPCR signaling predominantly alters viral infection in resting cells.

Figure 3

Inhibition of HIV-1 infection by CB2 activation is enhanced in CD4+ T cells infected prior to activation.

CB2R agonism inhibits SDF-1α mediated CXCR4 signaling

Our data suggested that CB2R activation specifically inhibited CXCR4-tropic virus and that this effect was greatest in resting cells. Given a previous report, which indicated that CXCR4 signaling enhances HIV-1 infection in resting cells , we chose to investigate the functional consequence of CB2R stimulation on CXCR4-mediated signaling. To assay for CXCR4 activity, we first measured changes in a G protein activity in CD4+ T cells activated by the cytokine SDF-1α (CXCL12). SDF-1α treatment led to a robust, dose-dependent increase in [35S]GTPγS binding in CD4+ T cells (Figure 4A). Pretreatment with 100 nM JWH-133 significantly decreased this effect at higher doses (16.7±8.37%, **p = 0.0034) (mean±SEM, n = 4 donors) (Figure 4B). In contrast, pretreatment with SDF-1α did not inhibit JWH-133 induced [35S]GTPγS binding at any concentration tested (Figure 4C). CB2 agonism therefore decreases G-protein activation in response to the CXCR4 agonist SDF-1α.

Figure 4

CB2 agonism inhibits SDF-1á mediated G-protein coupling to CXCR4.

The CB2R-induced inhibition of CXCR4 signaling was also indicated by a decrease in phosphorylation of downstream kinases following SDF-1α treatment (Figure 5). We assayed SDF-1α-induced Akt and p32/44 MAP kinase (ERK 1/2) phosphorylation in CD4+ T cells following treatment with or without CB2R agonist JWH-133. SDF-1α-mediated phosphorylation of both Akt (Figure 5A–B) and ERK 1/2 (Figure 5C–D) was significantly inhibited by pretreatment with JWH-133. These results suggest that the activation of CB2R is sufficient to inhibit downstream CXCR4-mediated signaling pathways.

Figure 5

CB2 agonism inhibits acute SDF-1á mediated signaling in primary CD4+ T cells.

The CB2R agonist JWH-133 decreases F-actin in CD4+ T cells

Down-regulation of CXCR4 signaling with pertussis toxin has been shown to decrease actin dynamics, disrupting the remodeling of the cortical actin barrier required for HIV-1 infection . Given the capacity for CB2R agonism to inhibit upstream CXCR4-mediated signaling events, we hypothesized that CB2R could suppress CXCR4-induced actin polymerization. To test for this possibility, we assayed the ability of JWH-133 to inhibit SDF-1α mediated actin polymerization, as visualized by phalloidin binding. While a significant difference in the rate of SDF-1α induced F-actin accumulation was not detected (data not shown), we did find that JWH-133-treatment led to a significant decrease in phalloidin binding as compared to untreated control cells (Figure 6A–B) (at time 0, 46.2±17.46 versus 64.36±24.34 MFI, respectively; after 30 minutes of SDF-1α treatment, 23.36±11.68 versus 31.3±15.65 MFI, respectively) (mean±SEM, n = 7, **p = 0.0021, *p = 0.04). Although CB2R agonism was found to disrupt SDF-1α induced G protein binding and signaling (Figure 4 and 5),5), this decrease did not translate into acute disruption of the rate of actin polymerization following SDF-1α treatment. Therefore it is possible that CB2R exerts its effect at the level of F-actin formation as treatment with JWH-133 reduces the level of F-actin in the steady state.

Figure 6

Alteration of actin organization in primary CD4+ T cells pretreated with CB2 agonist.

During HIV-1 infection, the virus itself acts as an agonist to stimulate CXCR4 and induce actin remodeling in resting cells . We tested to see whether changes in actin reorganization caused by viral binding were altered by CB2R agonist pretreatment. To do so, we pretreated CD4+ T cells with JWH-133, incubated these cells with HIV-1 viral particles, and then measured changes in phalloidin binding to F-actin over time. We found that HIV-1 induced transient upregulation in phalloidin binding activity peaking at approximately 1 minute after addition in both control and JWH-133 treated cells (Figure 6C). Like with SDF-1α treatment, the rate of increase in phalloidin binding was not significantly altered in CB2R agonist pretreated cells. Consistent with our previous observations, we found that that basal phalloidin binding was significantly reduced in JWH-133 treated cells. The reduction in F-actin formation in JWH-133 treated cells was consistent over time despite the transient upregulation of phalloidin stain induced by virus (Figure 6C). Taken together, these data suggest that reduction of F-actin induced by CB2R agonism results in the reduction of total F-actin over time, despite addition of CXCR4 agonists such as SDF-1α or HIV-1 viral particles. The rate of actin polymerization remained constant after acute CXCR4 activation, but the total amount of F-actin induced was significantly lower in cells pretreated with CB2R agonist. Indeed, the amount of F-actin induced by wild type virus in JWH-133 pretreated cells was similar to that induced by Env-null viral particles in control treated cells (at 1 minute, 107.5±53.7 versus 95.9±47.9) (mean±SEM, n = 4) (Figure 6D). Therefore, JWH-133 treatment reduces F-actin concentration to background levels, that is, the same level as non-specific induction of actin by Env-null virus. This basal reduction in actin polymerization via CB2R may reduce actin rearrangement sufficiently to block viral infection.

CB2R agonism decreases HIV-induced cofilin activation

HIV-triggered actin rearrangement is regulated in part by the severing protein cofilin, which dissociates and facilitates depolymerization of actin filaments thereby promoting actin dynamics . In the inactive state, cofilin is phosphorylated; agonism of Gαi-coupled chemokine receptors initiates cofilin de-phosphorylation and activity . Induction of cofilin de-phosphorylation increases HIV infection , we therefore hypothesized that JWH-133 treatment, which decreases HIV infection, would lead to increased cofilin phosphorylation. To identify changes in cofilin regulation, we measured phosphorylated cofilin (p-cofilin) expression in CD4+ T cells pretreated with JWH-133 and exposed to HIV. We detected a significant increase of p-cofilin over time in JWH-133 treated cells, as compared to control treated cells (216.42±124.95 versus 146.83±84.77, respectively, p-cofilin density normalized to actin at 120 minutes) (mean±SEM, n = 3, *p<0.05) (Figure 6E–F). This data suggests that CB2R agonism not only blocks HIV-induced cofilin de-phosphorylation, but also enhances p-cofilin. Indeed, CB2R-induced p-cofilin was also observed in cells exposed to a control, VSV-envelope pseudotyped virus (183.11±91.56 JWH-133 versus 138.32±69.16 control, p-cofilin density normalized to actin at 120 minutes) (mean±SEM, n = 4, *p<0.05) (data not shown). We did not observe changes in total cofilin levels in cells treated with HIV and JWH-133 (Figure 6G). Taken together, our data indicate that CB2R agonist pretreatment leads to accumulation of p-cofilin as well as the inhibition of cofilin dephosphorylation, i.e. activation in the presence of HIV; findings consistent with a known requirement for cofilin activation in HIV infection . The increase of inactive cofilin in JHW-133 treated cells is a likely mechanism for the reduction of actin dynamism and subsequent inhibition of viral infection in these cells.

CB2R agonism inhibits productive infection, but not viral fusion

Alteration of actin dynamics has been shown to inhibit productive viral infection but not viral binding or fusion . To confirm that CB2R-mediated changes in the actin cytoskeleton did not inhibit viral fusion, we assessed levels of fusion using a β-lactamase (BlaM)-Vpr fusion assay . NL4.3 virions that incorporated BlaM-VPR were used to infect JWH-133 and control treated cells loaded with the β-lactamase substrate CCF2-AM. Upon viral membrane fusion, BlaM-Vpr is released into the cytoplasm where it is able to cleave CCF2-AM. Infection of primary, resting CD4+ T cells induced fusion in approximately 4% of cells (Figure 7A). We found that blockade of HIV binding with the CXCR4 antagonist AMD3100 significantly reduced viral fusion and CCF2-AM cleavage, but pretreatment with JWH-133 had no effect (Figure 7 A–B). This data indicates that CB2R antagonism of CXCR4 function does not block the early stages of viral infection binding and fusion.

Figure 7

CB2 agonism inhibits productive infection of CXCR4-tropic HIV-1, but not cell-associated transfer or viral fusion.

CB2R agonism inhibits cell-associated viral infection, but not transfer

Our findings suggest that CB2R agonism strongly inhibits post-fusion events during viral infection in resting cells exposed to cell-free virus. Given these observations, we wanted to determine the capacity for CB2R agonism to block viral transmission and infectivity in a cell-associated model of infection. Cell-associated viral infection is hundreds to thousands-fold more efficient than cell-free infection . During cell-associated transmission, a synaptic structure, called the viral synapse, is formed between the infected (“donor”) and non-infected (“target”) cell . Significant actin rearrangements accompany formation of the virological synapse , and these actin structures have been hypothesized to regulate viral penetration into the target cell . Unlike cell-free viral infection, transfer of virus between cells is co-receptor independent; blockade of viral binding to CXCR4 with a selective antagonist, AMD3100, does not inhibit passage of virus . Once virus is captured within a target cell, however, co-receptor binding is still required for viral fusion. We hypothesized that CB2R agonism, like the CXCR4 antagonist AMD3100, would not inhibit viral transfer, but would block productive infection.

To assay the impact of CB2R agonism on cell-associated viral transfer, we used a CXCR4-tropic NL4-3-based reporter virus called HIV Gag-iGFP, which carries an interdomain insertion of green fluorescent protein (GFP) in the core structural protein Gag . Each mature viral particle contains ∼5000 GFP molecules(53), so viral transmission can be measured with high sensitivity . We pretreated CD4+ T cell targets with AMD3100 or JWH-133, and then co-cultured these cells with dye-labeled Jurkat donors infected with the HIV Gag-iCherry reporter virus. After three hours of co-culture, we assessed viral transmission to the target population (Figure 7C–D). As previously reported, we found no difference in expression of the HIV Gag-iCherry reporter virus in CD4+ T cells pretreated with AMD3100 as compared to control treated cells. Likewise, pretreatment with the CB2 agonist JWH-133 did not impair viral transfer into target cells (Figure 7D). Within this T cell population, transmission to memory (CD45RO+) CD4+ T cells was approximately 50% more efficient than transfer into naive (CD45RO-) cells. Transfer to both T cell subsets was as efficient in AMD3100 and JWH-133 cells as control treated cells. These findings confirm that CB2R-mediated inhibition of CXCR4 function does not impair cell-associated HIV-1 transfer to either naive or memory T cells.

We next sought to determine whether CB2R agonist pretreatment blocked productive infection following cell-to-cell transfer of virus. We pretreated cells with either AMD3100 or JWH-133 and then conducted a 3-hour transfer experiment using dye-labeled donor CD4+ T cells infected with the NL-GI virus, the same as was used for cell-free assessment of productive infection (Figure 1). We then sorted the naïve (CD45RO-) targets from the memory (CD45RO+) target T cell population and activated both subsets to initiate viral replication. We found that in both naive and memory T cells, both AMD3100 and JWH-133 pretreatment inhibited productive viral infection (Figure 7E–G). This inhibition was significant in the memory population, with CB2R agonist pretreatment resulting in an approximately 50% decrease in infected cells after four days of culture (4.02±2.0% versus 7.47±3.7% control) (mean±SEM, n = 4, *p<0.05) (Figure 7F). Although naive cells overall exhibited productive infection at a much lower frequency than memory cells, JWH-133 pretreatment reduced the number of infected cells (0.02±0.01% versus 0.2±0.1%) (mean±SEM, n = 4) (Figure 7E). These results indicate that CB2R agonism blocks productive viral infection after cell-associated viral exposure, just as it does with cell-free virus. The capacity for CB2R agonism to block infection following viral transfer is consistent with our finding that CB2R-mediated inhibition infection occurs after binding, and indeed fusion, in a cell-free system (Figure 7A–B). Taken together, these results are consistent with the idea that actin rearrangements inhibited by CB2R, while not required for cell-associated viral transfer, are required for productive viral infection following cell-associated transfer.


Human immunodeficiency virus type 1, (HIV-1) infection in T cells requires viral binding to two receptors, CD4+ and a chemokine co-receptor, either CXCR4 or CCR5 . These co-receptors are members of the highly conserved family A of G-protein coupled receptors (GPCRs). Absence of co-receptors, or blockade of HIV-1 binding to one of these co-receptors, are both sufficient to abrogate de novo viral infection in a target cell . Similarly, manipulation of these GPCRs with pharmacological ligands that alter co-receptor recycling , binding pocket occupancy , or co-receptor activity  also inhibit viral replication.

Here, we report that cannabinoid activation of CB2R inhibits CXCR4-tropic HIV infection by altering CD4+ T cell actin dynamics. We find that selective CB2 activation blocks both cell-free and cell-associated viral infection, reducing the frequency of infected cells by 30-60% (Figures 1,,7).7). This inhibition is pronounced in resting cells, which are a target of CXCR4-tropic HIV . Additionally, this inhibition is mediated post-transfer during cell-to-cell infection and post-fusion in the target cell following infection with cell-free virus (Figure 7). We further investigated the mechanism by which CB2R agonism altered HIV permissiveness. Our findings demonstrate that CB2R activation at concentrations sufficient to inhibit virus does not alter CXCR4 levels of surface expression, but does significantly reduce CXCR4-mediated G-protein binding and downstream signaling (Figure 4 and and5).5). This inhibition of CXCR4 signaling is accompanied by a loss in F-actin accumulation (Figure 6), which may prevent the cortical actin rearrangements required for reverse transcription and migration of the viral preintegration complex to the nucleus . Taken together, our results suggest that CB2 cross-regulates CXCR4 and that this inhibitory cross-talk is sufficient to decrease viral infection.

Although we here identify cross-talk between the CB2R and CXCR4 receptors and downstream impairment of actin dynamics, the possibility remains that CB2R activation results in induction of unknown anti-viral host factors. Arguing against the possible induction of unknown anti-viral factors, CB2R agonism did not block HIV infection by virus bearing the CCR5-tropic JRFL envelope (Figure 1C). The inability of CB2 to inhibit CCR5-tropic virus suggests that CB2-mediated alterations to the target cell are negligible in the predominantly memory CCR5+ CD4+ T cell subset. This effect is unlikely to be specific to memory cells as a whole, as CB2R treatment efficiently blocked productive infection in both memory and naive cell subsets following infection with cell-associated X4-tropic HIV-1 (Figure 7F–G). Rather, these findings may indicate that infection with CCR5-tropic virus is less dependent on chemokine-receptor mediated signaling and de novo cytoskeletal rearrangement for productive infection.

The CB2R may be considered as an adjunct therapeutic target for inhibition of CXCR4-tropic viral spread to resting T cell populations in patients with AIDS. A particularly compelling therapeutic rationale for the evaluation of antiviral affects of CB2R agonists may be to address severe symptoms of cachexia or neuropathic pain which may also present in patients with AIDS, without the adverse neurological or behavioral side effects associated with CB1R agonism . Although the effect of CB2R agonists on HIV infection is moderate, an accumulated effect in patients treated daily for pain could explain positive effects on viral load over time. We find that CB2R agonist pretreatment of resting cells inhibits viral spread in a receptor-selective manner in resting cells and does broadly inhibit T cell activation (Figure 2). Although previous studies have indicated that pan-cannabinoid agonists can possess an immunosuppressive function in vitro  and in vivo , ablation of CB2R in mice was not found to increase T cell number, proliferation, or apoptosis in the periphery . Immunosuppression by CB2R may be attributed in part to drug toxicity at high concentrations of cannabinoid agonist. Indeed, the use of cannabinoid drugs in patients with HIV is associated with an increase, rather than a decrease, in CD4+ T cell number  and has been shown to reduce viral load in SIV infected rhesus macaques . It is possible that novel CB2R-specific agonists and allosteric modulators that exert potent anti-viral activity without inducing immunosuppression could be identified. Further study of cannabinoids and other neuroendocrine regulators that selectively modulate immune function may result in the discovery of new anti-viral drugs that can also mitigate AIDS-associated symptoms.


The authors would like to thank I. Gomes, A. Del Portillo, R. Rosenfeld, R. Alvarez, and S. Stockton for constructive advice and technical support.


Competing Interests: The authors have declared that no competing interests exist.

Funding: This work was supported by United States National Institutes of Health grants R01AI074420 and DP1DA028866 awarded to BKC and NIH grants DA008863 and DA019521 awarded to LAD. CMC is further supported by NIH CTSA grant UL1RR029887 awarded to the Mount Sinai School of Medicine. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


Objective: To determine the effect of smoked cannabis on the neuropathic pain of HIV-associated sensory neuropathy and an experimental pain model.

Methods: Prospective randomized placebo-controlled trial conducted in the inpatient General Clinical Research Center between May 2003 and May 2005 involving adults with painful HIV-associated sensory neuropathy. Patients were randomly assigned to smoke either cannabis (3.56% tetrahydrocannabinol) or identical placebo cigarettes with the cannabinoids extracted three times daily for 5 days. Primary outcome measures included ratings of chronic pain and the percentage achieving >30% reduction in pain intensity. Acute analgesic and anti-hyperalgesic effects of smoked cannabis were assessed using a cutaneous heat stimulation procedure and the heat/capsaicin sensitization model.

Results: Fifty patients completed the entire trial. Smoked cannabis reduced daily pain by 34% (median reduction; IQR = −71, −16) vs 17% (IQR = −29, 8) with placebo (p = 0.03). Greater than 30% reduction in pain was reported by 52% in the cannabis group and by 24% in the placebo group (p = 0.04). The first cannabis cigarette reduced chronic pain by a median of 72% vs 15% with placebo (p < 0.001). Cannabis reduced experimentally induced hyperalgesia to both brush and von Frey hair stimuli (p ≤ 0.05) but appeared to have little effect on the painfulness of noxious heat stimulation. No serious adverse events were reported.

Conclusion: Smoked cannabis was well tolerated and effectively relieved chronic neuropathic pain from HIV-associated sensory neuropathy. The findings are comparable to oral drugs used for chronic neuropathic pain.



No studies to date have directly compared the tolerability and efficacy of smoked marijuana and oral dronabinol in HIV+ marijuana smokers.


The aim of this study was to compare dronabinol (0, 10, 20, 30 mg p.o.) and marijuana [0.0, 1.8, 2.8, 3.9% Δ9-tetrahydrocannabinol (THC)] in two samples of HIV+ marijuana smokers: those with (n=15) and those without (n=15) a clinically significant loss of muscle mass (<90% body cell mass/height), which is one component of AIDS wasting.


Mood, physical symptoms, self-selected food intake, cardiovascular data, and cognitive task performance were measured before and repeatedly after dronabinol and marijuana administration in eight 7-h sessions. Marijuana and dronabinol were administered in randomized order using a within-subject, staggered, double-dummy design.


As compared to placebo, (1) marijuana (1.8, 2.8, 3.9% THC) and the lower dronabinol doses (10, 20 mg) were well tolerated (e.g., few physical symptoms, significant increases in ratings of “good drug effect”) in both groups of participants; the highest dose of dronabinol (30 mg) was poorly tolerated in a subset of participants; (2) marijuana and dronabinol significantly increased caloric intake in the low bioelectrical impedance analysis (BIA) group but not in the normal BIA group; and (3) drug effects on cognitive performance were minor.


These data suggest that for experienced marijuana smokers with clinically significant muscle mass loss, both dronabinol (at acute doses at least four to eight times the current recommendation) and marijuana produce substantial and comparable increases in food intake without producing adverse effects.

EQUAL MEDICINE ORGANIZATION - Marijuana ALS Amyotrophic Lateral Sclerosis Treatment cures medical marijuana thc cannabis cbd