Content uploaded by Renata Marchette
Author content
All content in this area was uploaded by Renata Marchette on Jan 13, 2023
Content may be subject to copyright.
Addiction Neuroscience 5 (2023) 100057
Contents lists available at ScienceDirect
Addiction Neuroscience
journal homepage: www.elsevier.com/locate/addicn
Extended access to fentanyl vapor self-administration leads to
addiction-like behaviors in mice: Blood chemokine/cytokine levels as
potential biomarkers
Renata C.N. Marchette
a
,
∗
, Erika R. Carlson
a
, Nadia Said
a
, George F. Koob
a
,
Leandro F. Vendruscolo
b
a
Neurobiology of Addiction Section, National Institute on Drug Abuse, Intramural Research Program, National Institutes of Health, 251 Bayview Boulevard,
BRC 08A505.16, Baltimore, MD 21224, USA
b
Stress & Addiction Neuroscience Unit, National Institute on Drug Abuse and National Institute on Alcohol Abuse and Alcoholism Intramural Research Programs,
National Institutes of Health, Baltimore, MD, USA
Keywords:
Substance use disorder (SUD)
Opioid use disorder (OUD)
Hyperalgesia
Extended access
Addiction-like behavior
Operant self-administration
Punishment
Rodent models are useful for understanding the mechanisms that underlie opioid addiction, but most preclinical
studies have focused on rewarding and consummatory aspects of opioids without components of dependence-
induced escalation of drug taking or seeking. We characterized several opioid-related behaviors in mice using
a model of vaporized fentanyl self-administration. Male and female C57BL/6J mice were assigned to short-
access (ShA; 1 h, nondependent) or long-access (LgA; 6 h, dependent) fentanyl vapor self-administration and
subsequently tested in a battery of behavioral tests, followed by blood collection during withdrawal. Com-
pared with mice in the ShA group, mice in the LgA group escalated their fentanyl intake, were more mo-
tivated to work to obtain the drug, exhibited greater hyperalgesia, and exhibited greater signs of naloxone-
precipitated withdrawal. Principal component analysis indicated the emergence of two independent behav-
ioral constructs: “intake/motivation ”and “hyperalgesia/punished seeking. ”In mice in the LgA condition only,
“hyperalgesia/punished seeking ”was associated with plasma levels of proinammatory interleukin-17 (IL-17),
chemokine (C-C motif) ligand 4 (CCL-4), and tumor necrosis factor 𝛼(TNF- 𝛼). Overall, the results suggest that ex-
tended access to opioids leads to addiction-like behavior, and some constructs that are associated with addiction-
like behavior may be associated with levels of the proinammatory cytokines/chemokines IL-17, TNF- 𝛼, and
CCL-4 in blood.
1. Introduction
Opioid use disorder (OUD) is a chronic relapsing disorder that
is characterized by a pattern of opioid use that leads to clini-
cally signicant impairment and distress [1] . In 2019, 1.2% of
the population worldwide used opioids, and opioids were present
in over 70% of overdose deaths [2] . The nonmedical use of opi-
oids is a major problem in North America but has remained sta-
ble since 2010. Opioid overdose deaths doubled in the same pe-
riod, suggesting an increase in the nonmedical use of potent opi-
oids, such as fentanyl and its analogs [2] . During the coronavirus
disease 2019 (COVID-19) pandemic, the United States Centers for
Disease Control and Prevention reported a 35% increase in opioid
overdose deaths ( https://www.cdc.gov/media/releases/2020/p1218-
overdose-deaths-covid-19.html ).
∗ Corresponding author.
E-mail address: renata.marchette@nih.gov (R.C.N. Marchette) .
Opioid use disorder involves a compulsion to seek and take opioids,
the loss of control over intake, and the emergence of a negative aect
state during withdrawal, also known as hyperkatifeia, which are symp-
toms that reect multiple sources of motivation for opioid seeking [3] .
There are also physical symptoms, such as body aches, diarrhea, greater
pain sensitivity, and lower pain tolerance (hyperalgesia) during opioid
withdrawal [ 4 , 5 ]. The treatment of OUD is largely limited to drugs that
act on the opioid system (e.g., µ-opioid receptor agonist methadone, µ-
opioid receptor partial agonist buprenorphine, and preferential µ-opioid
receptor antagonist naltrexone). The preferential µ-opioid receptor an-
tagonist naloxone (Narcan®) is used to reverse opioid overdose. Treat-
ment with the 𝛼2
-adrenergic receptor agonists lofexidine and clonidine
is restricted as an adjuvant during medically supervised withdrawal
[6] . Adherence and retention to the medication-assisted treatment of
OUD faces several barriers, such as poor accessibility, high cost, and
https://doi.org/10.1016/j.addicn.2022.100057
Received 24 August 2022; Received in revised form 7 December 2022; Accepted 8 December 2022
2772-3925/Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
stigma [ 6 , 7 ]. Additionally, chronic treatment with buprenorphine and
methadone may not improve hyperalgesia [8] , which may contribute to
the maintenance of drug taking and relapse. Therefore, understanding
the biological mechanisms that underlie motivational withdrawal and
contribute to drug seeking and taking may shed light on OUD etiology
and assist with medication development.
Animal models are a key tool for understanding the neurocircuitry
that underlies OUD [ 9 , 10 ] and testing novel pharmacological targets
[ 11 , 12 ]. One of these models involves extended access to intravenous
opioid self-administration, which recapitulates several characteristics
of OUD. Rats [ 13 , 14 ] and mice [15] escalate their heroin intake, ex-
hibit greater signs of naloxone-induced withdrawal, and are more mo-
tivated for the drug when given extended access (e.g., 6–24 h self-
administration sessions) to the drug compared with limited access
(e.g., 1–3 h self-administration sessions). Although intravenous self-
administration models are considered the gold-standard in addiction
research, they pose several technical challenges, such as high rates of
catheter failure, especially in mice [ 16 , 17 ]. Considering the importance
of fentanyl and its analogs in the current opioid crisis, we developed va-
porized fentanyl and sufentanil self-administration models in both rats
[ 18 , 19 ] and mice [20] that produce motivational and physical (somatic)
signs of opioid withdrawal without the need for a catheter implant. Ad-
ditionally, most research has focused on reward and consummatory be-
haviors without components of dependence-induced increases in drug
taking and seeking, and drug taking and seeking despite punishment
(i.e., addiction-like behaviors, [21] ). Whether extended access to fen-
tanyl vapor self-administration leads to addiction-like behaviors in mice
remains to be determined.
Addiction-like behaviors are characterized by persistent, repetitive
drug seeking and taking that can prevent or provide relief from distress,
anxiety, or stress [ 1 , 22 , 23 ]. For the purposes of the present study, el-
ements of addiction-like drug seeking were divided into the following
phenotypic components: (i) escalation of drug intake, (ii) increase in
drug seeking under progressive-ratio (PR) and progressive-delay condi-
tions, and (iii) drug seeking and taking despite aversive consequences
[24] .
Neuroimmune alterations in the brain have been shown to signi-
cantly alter key neurotransmitters involved in substance use disorders
[25] . As such, the overproduction of cytokines and other immune sig-
naling molecules can decrease dopamine and increase glutamate sig-
naling and release (for review, see [26–28] ). Conversely, prolonged ex-
posure to an addictive drug aects the neuroimmune system by elic-
iting microglia and astrocyte activation, leading to the release of cy-
tokines and chemokines [29] . Opioids induce many neuroinamma-
tory alterations in the brain that may contribute to the development
of addiction through myriad mechanisms [29] . Opioids activate mi-
croglia (i.e., resident immune cells in the brain) and astrocytes (i.e.,
neuronal regulatory cells; [30] ) leading to the release of proinamma-
tory cytokines [31] . Recent studies reported an upregulated expression
of cytokines such as TNF- 𝛼, IL-1 𝛽, IL-6 and IL-17 [32] . Opioids can also
increase the production of chemokines like CCL2 and CCL4 in human
neurons [33] and astrocytes [34] . However, the role of peripheral cy-
tokines and chemokines as biomarkers of addiction-like opioid intake
remains to be investigated. Therefore, we chose to explore the impact
of prolonged opioid exposure on those specic pro-inammatory cy-
tokines (TNF- 𝛼, IL-1 𝛽, IL-6, IL-17) and chemokines (CCL2, CCL4) in
blood.
Using a battery of behavioral tests, the present study investigated
opioid dependence-related behaviors, including drug seeking and tak-
ing despite punishment, in the vaporized fentanyl self-administration
model in male and female mice. The secondary goal was to assess blood
levels of cytokines and chemokines as potential biomarkers of opioid
dependence. Our hypothesis was that extended access to fentanyl vapor
self-administration will result in qualitative and quantitative changes in
opioid-related behaviors and that blood cytokine and chemokine levels
serve as biomarkers for some of these behaviors.
2. Methods
2.1. Animals
C57BL/6J mice (41 females and 41 males) were purchased from
Jackson Laboratories (Bar Harbor, ME, USA) at 7 weeks of age and were
8–10 weeks old at the beginning of the experiment. The mice were kept
in groups of two to four per cage and housed under a reverse light cy-
cle (lights on at 7 PM) with controlled temperature (22 °C ± 2 °C) and
humidity (50–60%). The mice had free access to water and food except
during testing. Tests were performed during the dark cycle. Body weight
was recorded at least weekly. All procedures followed the National In-
stitutes of Health Guide for the Care and Use of Laboratory Animals
and were previously approved by the National Institute on Drug Abuse
Animal Care and Use Committee.
2.2. Drugs
Fentanyl citrate (Mallinckrodt, St. Louis, MO, USA) was dispensed by
the National Institute on Drug Abuse Intramural Research Program phar-
macy. The stock solution was prepared by adding 6 mL of sterile water
to 1 g fentanyl and 44 mL of vehicle (80% vegetal glycerin [VG]/20%
propylene glycol [PG]). The stock solution (20 mg/mL) was diluted with
vehicle to a concentration of 5 mg/mL for the self-administration ex-
periments [20] . Capsaicin-adulterated fentanyl was used for drug seek-
ing and taking despite punishment. Capsaicin (catalog no. N735–5 g,
AK Scientic, Union City, CA, USA) was dissolved in 0.5 mL of 200-
proof ethanol (catalog no. 111,000,200, Pharmco-Aaper, Brookeld, CT,
USA). This solution was then added to the vehicle (VG/PG) or 5 mg/mL
fentanyl in VG/PG. The mixture was sonicated at 40 °C for 1 h.
2.2.1. Fentanyl and norfentanyl measurement in blood
All mice received 1, 3, or 10 noncontingent vapor deliveries (1.5 s,
60 W) of fentanyl, evenly spaced (for three and 10 deliveries) in 30 min
( Fig. 1 A). Immediately after the end of the session (2 min after the last
fentanyl delivery), the mice were deeply anesthetized in a chamber that
was saturated with isourane and euthanized by decapitation. Trunk
blood was collected in microtubes that contained 45 µL of 4% ethylene-
diaminetetraacetic acid (EDTA). Blood samples were analyzed by high-
performance liquid chromatography/tandem mass spectrometry at NMS
Labs (Horsham, PA, USA).
2.3. Nociception tests
To measure antinociception, we used a hot plate (Ugo Basile, Gemo-
nio, Italy) at 52.5 °C. The latency to the rst nociceptive behav-
ior was recorded at baseline immediately before an acquisition self-
administration session (BL) and immediately after an acquisition self-
administration session (ACQ). Nociceptive behaviors included jumps
and rear paw licking or vigorous icking. A cuto time of 30 s was im-
posed to prevent tissue damage. To measure mechanical hyperalgesia
during withdrawal, we used an electronic von Frey device (Ugo Basile,
Gemonio, Italy). All mice were tested before acquisition session 1 for
their baseline responding and immediately before escalation session 10
(i.e., 36 h after the end of their last self-administration session). The
maximum force was set to 50 gf, and the paw withdrawal threshold
was determined as the force necessary to elicit a response (i.e., paw re-
moval). Two measures were taken for each rear paw with an interstimu-
lus interval of at least 30 s. We calculated the average of these measures
and used this as a measure of mechanical sensitivity. For principal com-
ponent analysis (PCA), this value was multiplied by − 1, so higher values
represented more hyperalgesia.
2.4. Fentanyl vapor self-administration
The fentanyl vapor self-administration operates in a closed system.
The vaporizer only works with the chamber closed in an airtight envi-
2
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
Fig. 1. Fentanyl vapor leads to detectable blood fentanyl and norfentanyl levels and antinociception. (A) Temporal distribution of noncontingent fentanyl
vapor deliveries. (B) Concentration of fentanyl and norfentanyl in blood in male and female mice. Eighteen mice were exposed to 1, 3, or 10 fentanyl vapor deliveries
(1.5 s, 60 W) and euthanized 2 min after the last vapor exposure for blood collection. Fentanyl and norfentanyl levels were analyzed by liquid chromatography/tandem
mass spectroscopy. The data are expressed as individual points and were analyzed by linear regression ( n = 6/number of deliveries). (C) Antinociceptive response
to fentanyl vapor self-administration. Male and female C57BL/6J mice were tested in the hot plate test (52.5 °C, 30 s cuto) immediately before and after the fth
1-h fentanyl vapor self-administration acquisition session. Fentanyl vapor self-administration increased the latency to a nociceptive response. The data are expressed
as the mean ± SEM and were analyzed using paired Student’s t -test.
∗ ∗ ∗ ∗
p < 0.0001. Antinociceptive responses positively correlated with the number of fentanyl
vapor deliveries in the self-administration session. The data are expressed as individual points and were analyzed by linear regression. BL, baseline; ACQ, acquisition
( n = 16 mice/sex).
ronment. The pump produces negative pressure to pull fentanyl vapor
inside the chambers. The outlet tubes connect to a HEPA lter before the
vapor is released to the building exhaust system. All procedures were
performed as previously described [35] . Briey, mice were trained to
lever press for vaporized fentanyl (1.5 s, 60 W, 1 min timeout) on a
xed-ratio 1 (FR1) schedule of reinforcement, in which each lever press
on the active (left) lever resulted in vapor delivery. The concentration
of fentanyl used was determined in our previous work [20] . Drug de-
livery was accompanied by activation of a cue light. The cue light re-
mained on during the timeout period to signal that the drug was not
available. Presses on the inactive (right) lever and during the timeout
period were recorded but had no consequence. After acquisition (six 1-h
sessions) of fentanyl vapor self-administration, the mice were split into
two groups. One group was allowed 1 h access (short access [ShA]), and
the other group was allowed 6 h access (long access [LgA]) to fentanyl
self-administration in 10 FR1 sessions (escalation phase). The mice were
then tested for motivation to obtain fentanyl under PR and delay sched-
ules (described below), nociception, and capsaicin-punished drug seek-
ing. After 3 weeks of abstinence, all mice underwent 10 re-escalation
sessions, after which they were tested for naloxone-precipitated with-
drawal and capsaicin-punished drug taking (described below). An FR1
session following the same parameters described above was performed
between tests to maintain stable levels of lever pressing in the ShA group
and opioid dependence in the LgA group.
2.5. Progressive-ratio test
We tested the mice in a PR task, in which the number of lever presses
that was required for vapor delivery was sequentially increased by six
(PR 6; i.e., 1, 7, 13, 19, 25, 31, etc.). A 30 min period without vapor de-
livery or a total of 6 h ended the session. The breakpoint was determined
as the last ratio ( “eort ”) that was completed in the session.
2.6. Delay
We also tested the mice in a delayed-reward task, in which the in-
terval between lever pressing and vapor delivery was sequentially in-
creased by 6 s (i.e., 1, 7, 13, 19, 25, 31, etc.) for each subsequent drug
delivery. A 30 min period without vapor delivery or a total of 6 h ended
the session. The breakpoint was determined as the last ratio (in seconds;
“time ”) that was completed in the session.
2.7. Precipitated withdrawal test
Immediately after the end of the FR1 self-administration session,
each mouse received an intraperitoneal (IP) injection of naloxone
(1 mg/kg, 10 mL/kg) and were placed in a transparent box in front
of a mirror at a 45° angle for the improved visibility of behavior. The
observation and recording period lasted 20 min. The videos were ana-
lyzed by an experienced scorer who was blind to group assignment. The
number of paw tremors ( “clapping ”), jumps, and wet-dog shakes were
recorded, and each occurrence was attributed one point. The presence
of other “physical ” signs, such as diarrhea, genital grooming, abnormal
posture, ptosis, salivation, and teeth chattering, were also recorded, and
attributed one point each, regardless of frequency. A precipitated global
withdrawal score was calculated as the sum of points [ 36 , 15 ].
2.8. Drug seeking and self-administration in the presence of an aversive
stimulus
To model punished drug seeking, we tested the mice for the self-
administration of capsaicin alone (i.e., vehicle without fentanyl). The
mice were exposed to four concentrations of capsaicin (0, 0.01, 0.03,
and 0.1%, w/v) in VG/PG in 1-h sessions for both ShA and LgA groups.
The presentation of capsaicin followed a Latin-square design, 24 h apart,
to avoid the confound of extinction-like behavior. To model punished
drug taking, the fentanyl solution (5 mg/mL) was adulterated with in-
creasing concentrations of capsaicin. The concentrations of capsaicin
followed a log scale, starting at 0.01% (capsaicin in fentanyl [w/v]) up
to 3%. The sessions with capsaicin-adulterated fentanyl lasted 1 h for
both ShA and LgA and were performed 24 h apart in an ascending order
of concentration.
3
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
2.9. Plasma collection
After the end of the behavioral test battery, all mice underwent one
nal FR1 self-administration session (1 h for ShA and 6 h for LgA). Af-
ter 36 h, the mice were deeply anesthetized with isourane and decap-
itated. Trunk blood was collected in microtubes that contained EDTA.
Blood samples were centrifuged at 10,000 ×g at 4 °C for 10 min. Plasma
was collected and stored at − 80 °C until use.
2.10. Corticosterone measurement
Plasma samples were diluted 1:100 with assay buer, and the
enzyme-linked immunosorbent assay (ELISA) was performed according
to the manufacturer’s instructions (catalog no. ab108821, Abcam, Cam-
bridge, UK).
2.11. Cytokine and chemokine measurement
Plasma samples were diluted by a factor of 1.5 and analyzed using
an Ella microuid multiplex cartridge according to the manufacturer’s
instructions (ProteinSimple, San Jose, CA, USA).
2.12. Data analysis
After conducting analyses of variance (ANOVAs) for possible sex in-
teractions and nding none (see below), we reanalyzed the male and
female data combined using two-way repeated-measures ANOVA, fol-
lowed by Duncan’s post hoc test, or paired or unpaired Student’s t -test
as appropriate. All data separated by sex are presented in the Supple-
mental Material. The data are expressed as the mean ± standard error of
the mean (SEM). Linear regression was used to determine correlations
between nociception and fentanyl vapor deliveries. Values of p < 0.05
were considered statistically signicant. To assess relationships among
variables, we used Varimax-normalized principal components analysis
(PCA) with Eigenvalues greater than 1. Loading factors ≥ 0.6 were con-
sidered. We arbitrarily selected loading factors higher than 0.6 to be
conservative with the representation of variables in each factor. The
larger a loading’s relative magnitude, the more the variables are posi-
tively correlated with each other (positive values) or negatively corre-
lated with each other (negative values). We then extracted the individ-
ual loading values of each mouse and used Pearson’s test to determine
correlations between PCA values and blood chemokines/cytokines and
corticosterone levels (two-tailed comparison, 𝛼= 1%). One female ShA
mouse was not included in the cytokine and corticosterone analysis be-
cause of technical issues during plasma sample processing.
3. Results
3.1. Sex differences
We analyzed all data for sex eects (Supplemental Tables S1–4). We
employed three-way ANOVAs and found no sex ×group interactions for
any of the variables analyzed. An overall eect of sex on punished seek-
ing was found ( F
1,28
= 17.026, p = 0.0003; F > M ) , with no interaction
between sex and any other variable (Supplemental Tables S1, S3). Be-
cause no sex ×group interactions were found, we collapsed the male
and female data for subsequent analyses.
3.2. Fentanyl vapor leads to detectable blood fentanyl levels and
antinociception
We previously reported that vaporized fentanyl and sufentanil leads
to concentration-dependent levels of fentanyl and sufentanil in blood
[ 18 , 20 ]. Here, we conrmed that cumulative exposure to fentanyl vapor
led to higher blood levels of fentanyl ( F
1,16
= 9.888, p = 0.006) and its
metabolite, norfentanyl ( F
1,16
= 32.79, p < 0.001), in male and female
mice ( Fig. 1 B). One vapor delivery of fentanyl 5 mg/mL led to a mean
plasma level of 4.55 ng/mL, whereas 10 vapor deliveries resulted in an
average plasma concentration of 28.63 ng/mL.
Next, we trained 32 mice (16 male and 16 female) to lever press
for fentanyl vapor delivery. Upon stable active lever pressing ( ∼5–6
sessions), we tested their antinociceptive response immediately after
the end of the session in the hot plate test ( Fig. 1 C). Fentanyl vapor
self-administration caused an antinociceptive response ( t
31
= 6.210, p
< 0.0001) that positively correlated with the number of lever presses
(R
2 = 0.28, F
1,30
= 12.13, p = 0.0015). This indicates that mice will
self-administer fentanyl to levels of biological eect (i.e., antinocicep-
tion).
3.3. Escalation of fentanyl vapor self-administration in dependent mice
For the rst hour of the session, the two-way ANOVA showed a signif-
icant group ×session interaction ( F
9,270
= 3.880, p = 0.00012; Fig. 2 A).
The Duncan post hoc test showed that mice in the LgA group had more
vapor deliveries in session 4 ( p = 0.01), session 5 ( p < 0.0001), session
6 ( p < 0.0001), session 7 ( p = 0.01), session 9 ( p = 0.0007), and session
10 ( p < 0.0001) compared with session 1 and more vapor deliveries in
session 5 ( p = 0.02) and session 10 ( p = 0.006) compared with mice in
the ShA group. For the total 6 h session, the two-way repeated-measures
ANOVA showed a signicant group ×session interaction for the number
of vapor deliveries ( F
9,270
= 2.917, p = 0.0026; Fig. 2 B). The Duncan post
hoc test showed that mice in the LgA group had more vapor deliveries
in session 5 ( p = 0.03), session 8 ( p = 0.02), session 9 ( p = 0.01), and
session 10 ( p < 0.0001) compared with session 1. There were no dier-
ences in vapor deliveries across sessions in the ShA group ( Fig. 2 A, B).
Using linear regression, we calculated the slope of the escalation curve
for each mouse. Mice in the LgA group had signicantly higher slopes
than mice in the ShA group, further conrming their escalation of intake
( t
30
= 2.259, p = 0.03; Fig. 2 C). For the rst hour of re-escalation, the
two-way repeated-measures ANOVA did not show a signicant group x
session interaction ( F
9,270
= 1.880, p = 0.055; Fig. 2 D) nor main eect
of group ( F
1,30
= 1.657, p = 0.21) but showed a main eect of session
( F
9,270
= 3.473, p = 0.0005). Duncan’s post hoc showed that the overall
number of vapor deliveries was higher on sessions 7 ( p = 0.00008), 8
( p = 0.0001) and 10 ( p = 0.02) compared with session 1. The two-way
repeated-measures ANOVA did not show a signicant group ×session
interaction for the number of vapor deliveries for the total 6 h dur-
ing re-escalation ( F
9,270
= 1.567, p = 0.12; Fig. 2 E) but showed main
eects of group ( F
1,30
= 77.029, p < 0.0001; ShA < LgA) and session
( F
9,270
= 2.115, p = 0.03). Duncan’s post hoc test showed that the over-
all number of vapor deliveries was higher in session 8 than in session
1 ( p = 0.012). The calculated slopes for the re-escalation did not dier
between ShA and LgA conditions ( t
30
= 0.48, p = 0.63; Fig. 2 F). Mice
allowed LgA escalated their fentanyl intake during the escalation phase
and maintained higher levels of fentanyl intake but did not escalate fur-
ther on the re-escalation phase.
3.4. Increases in naloxone-precipitated withdrawal signs, hyperalgesia
during spontaneous withdrawal, and motivation for fentanyl
Unpaired Student’s t -test showed that mice in the LgA group ex-
hibited more signs of naloxone-precipitated withdrawal compared with
mice in the ShA group ( t
30
= 3.424, p = 0.002; Fig. 3 A). Unpaired Stu-
dent’s t -test showed that mice in the LgA group had lower thresholds
than mice in the ShA group in response to mechanic stimulation with
an electronic von Frey device, indicating the development of hyperalge-
sia ( t
30
= 2.409, p = 0.049; Fig. 3 B). Unpaired Student’s t -test showed
that mice in the LgA group had a higher breakpoint than mice in the ShA
group in the PR test ( t
30
= 2.299, p = 0.03; Fig. 3 C). Unpaired Student’s
t -test showed that mice in the LgA group had a higher time breakpoint
than mice in the ShA group in the delay task ( t
30
= 2.090, p = 0.04;
Fig. 3 D). In summary, mice allowed LgA to fentanyl self-administration
4
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
Fig. 2. Escalation and re-escalation of fentanyl vapor self-administration. Male and female C57BL/6J mice were trained to lever press for fentanyl vapor
deliveries on an FR1 schedule of reinforcement. Upon stable lever pressing, the mice were split in short-access (ShA; 1 h) and long-access (LgA; 6 h) conditions. (A)
Mice in the LgA condition escalated their intake in the rst hour of the session across days. The data are expressed as the mean ± SEM and were analyzed using
two-way repeated-measures ANOVA.
#
p < 0.05, compared with session 1;
∗
p < 0.05, compared with LgA. (B) Mice in the LgA condition escalated their fentanyl intake
across the 6 h sessions. The data are expressed as the mean ± SEM and were analyzed using two-way repeated-measures ANOVA.
#
p < 0.05, compared with session
1;
∗
p < 0.05, compared with LgA. (C) Mice in the LgA condition had higher calculated slopes of the escalation curve compared with mice in the ShA condition. The
data are expressed as the mean ± SEM and were analyzed using unpaired Student’s t -test.
∗
p < 0.05, compared with ShA. (D) Fentanyl intake on the rst re-escalation
self-administration sessions. The data are expressed as the mean ± SEM and were analyzed using two-way repeated-measures ANOVA.
#
p < 0.05, compared with
session 1 . (E) Mice in the LgA condition maintained higher fentanyl intake across sessions in the re-escalation phase. The data are expressed as the mean ± SEM and
were analyzed using two-way repeated-measures ANOVA.
#
p < 0.05, compared with session 1;
∗
p < 0.05, compared with LgA. (F) The calculated re-escalation slope
did not dier between ShA and LgA conditions. The data are expressed as the mean ± SEM and were analyzed using unpaired Student’s t -test. ( n = 8/sex/group).
showed more signs of somatic withdrawal, more hyperalgesia and were
more motivated to obtain fentanyl than the ShA group.
3.5. Self-administration despite punishment
For punished fentanyl seeking, the two-way repeated-measures
ANOVA showed no dierence between groups ( p > 0.05) and no
group ×capsaicin concentration interaction ( p > 0.05) but a signicant
eect of capsaicin concentration ( F
3,90
= 12.60 , p < 0.0001; Fig. 4 A), in
which the self-administration of 0.03% and 0.1% capsaicin was lower
than vehicle without capsaicin (i.e., 0%). For punished taking, the two-
way repeated-measures ANOVA showed a signicant group ×capsaicin
concentration interaction ( F
6,180
= 3.797, p = 0.0014; Fig. 4 B). The post
hoc comparisons did not indicate a dierence between ShA and LgA
groups at any concentration of capsaicin-adulterated fentanyl. However,
in the ShA group, responding for 1% capsaicin ( p = 0.03) and 3% cap-
saicin ( p = 0.001) was signicantly lower than 0% capsaicin. In the LgA
group, responding for 0.3% capsaicin ( p = 0.006), 1% capsaicin ( p <
0.0001), and 3% capsaicin ( p < 0.0001) was signicantly lower than
vehicle. These data indicate that mice in both the ShA and LgA groups
reduced their drug seeking and taking in the presence of capsaicin.
3.6. Principal component analysis
To elaborate multidimensional phenotypes that are associated with
LgA to fentanyl compared with ShA, we performed a PCA with all behav-
ioral variables, combining males and females. We used the average of
the last three escalation sessions (i.e., sessions that were dierent from
session 1), the average of the 10 re-escalation sessions, the average of
0.03% and 0.1% capsaicin in vehicle without fentanyl, and the average
of fentanyl that was adulterated with 0.3%, 1%, and 3% capsaicin. We
analyzed the ShA and LgA groups separately. We did not restrict the
Varimax-normalized PCA to a pre-specied number of factors. Factor
loadings ≥ 0.6 were considered ( Fig. 5 ). For ShA, factor 1 represented
42.26% of total variance of the data and comprised positive associa-
tions between fentanyl intake (escalation and re-escalation) and moti-
vation (PR and delay). Factor 2 comprised 22.73% of the data variance
and included positive associations between hyperalgesia, capsaicin-
5
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
Fig. 3. Naloxone-induced signs of withdrawal, hyperalgesia during spontaneous withdrawal, and motivation for fentanyl . After escalation, male and female
C57BL/6J mice were tested in a battery of tests to assess somatic and motivational signs of withdrawal. (A) Naloxone-precipitated withdrawal. Immediately after
fentanyl vapor self-administration session 10, all mice received naloxone (1 mg/kg, IP) and were observed for 20 min for signs of withdrawal. The data are expressed
as the mean ± SEM and were analyzed using unpaired Student t -test.
∗ ∗
p < 0.01, compared with ShA. (B) Mechanical hyperalgesia. Between 36 to 40 h after a
self-administration session (i.e., during spontaneous withdrawal), the mice were tested for mechanical hyperalgesia using an electronic von Frey device. The data are
expressed as the mean ± SEM and were analyzed using unpaired Student t -test.
∗
p < 0.05, compared with ShA. The dotted line represents the average baseline measure
(i.e., before fentanyl exposure) for all mice; both groups developed hyperalgesia compared with the baseline (BL) measure. (C) Progressive ratio test (motivation or
“eort ”). After escalation, all mice were tested in a progressive-ratio task, in which the number of lever presses that were required for the next fentanyl delivery
increased by 6. The data are expressed as the mean ± SEM and were analyzed using unpaired Student’s t -test.
∗
p < 0.05, compared with ShA. (D) Time delay task
(motivation). After escalation, all mice were tested in the delayed-reward task, in which the time between lever presses and vapor delivery increased by 6 s. The
data are expressed as the mean ± SEM and were analyzed using unpaired Student’s t -test.
∗
p < 0.05, compared with ShA ( n = 8/sex/group).
Fig. 4. Self-administration despite punishment . Male and female C57BL/6J mice were tested for vapor self-administration in 1 h-sessions with vehicle or fentanyl
that was adulterated with increasing concentrations of capsaicin. (A) Vapor deliveries for each concentration of capsaicin alone in vehicle without fentanyl. The
data are expressed as the mean ± SEM and were analyzed using two-way repeated-measures ANOVA.
#
p < 0.05, compared with 0% regardless of group. (B) Vapor
deliveries for each concentration of capsaicin in fentanyl. The data are expressed as the mean ± SEM and were analyzed using two-way repeated-measures ANOVA.
#
p < 0.05, compared with 0% ( n = 8/sex/group).
6
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
Fig. 5. Principal component analysis . The behavioral measures were analyzed separately for each group (ShA and LgA) by a Varimax-normalized PCA. Factor
loadings ≥ 0.6 were considered. Black symbols represent all measures that loaded on factor 1 (x-axis). Orange circles represent measures that loaded on factor 2
(y-axis). The blue circle represents the only measure that loaded on factor 3. SWD, precipitated (somatic) withdrawal; PR, progressive ratio; vF, von Frey mechanical
nociception ( n = 8/sex/group).
punished seeking and taking, and naloxone-precipitated withdrawal.
For the LgA group, factor 1 (42.42% of the variance) showed positive as-
sociations between fentanyl intake (escalation and re-escalation), moti-
vation (PR and delay), and capsaicin-punished taking. Factor 2 (19.56%
of the variance) showed a positive association between hyperalgesia
and capsaicin-punished seeking. Factor 3 (14.84% of the variance) com-
prised naloxone-precipitated withdrawal signs. These data suggest that
there are qualitative dierences on the behavioral phenotype of mice
allowed ShA or LgA to fentanyl. Although PR and delay measures (mo-
tivation) were associated with FR1 (intake) in both groups, drug taking
despite punishment was associated with FR1 in the LgA group only. Only
in the ShA group hyperalgesia and naloxone-induced withdrawal were
part of the same behavioral construct.
3.7. Plasma levels of cytokines and corticosterone
There were no signicant group dierences in plasma corticosterone
levels ( p > 0.05), or cytokine and chemokine levels ( p > 0.05), including
iIL-1 𝛽, IL-6, IL-17, CCL-2, CCL-4, and TNF- 𝛼( Table 1 ).
After the behavioral test battery, the mice underwent an FR1 self-
administration session and were euthanized 36 h later (i.e., the time
point at which they would have undergone the next self-administration
session during spontaneous withdrawal). Trunk blood was collected, and
plasma was separated for corticosterone and cytokine analysis. The data
are expressed as the mean ± SEM and were analyzed using unpaired
Student’s t -test ( n = 7–8/sex/group).
Table 1
Plasma levels of corticosterone and cytokines.
ShA LgA
Corticosterone (ng/mL) 135.7 ± 19.85 152.3 ± 21.79
CCL-2 (pg/mL) 26.67 ± 3.15 23.17 ± 1.76
CCL-4 (pg/mL) 21.49 ± 1.35 23.15 ± 1.89
IL-1 𝜷(pg/mL) 1.87 ± 0.73 0.83 ± 0.15
IL-6 (pg/mL) 2.02 ± 0.84 1.03 ± 0.25
IL-17 (pg/mL) 4.14 ± 0.45 2.85 ± 0.39
TNF- 𝜶(pg/mL) 1.75 ± 0.10 1.49 ± 0.13
3.8. Correlation analysis
We extracted values of the principal components for each mouse
and correlated these values with plasma levels of cytokines/chemokines
and corticosterone using Pearson’s correlation ( Fig. 6 and Supplemen-
tal Table S5). We did not observe any correlations for the ShA group
( Fig. 6 A). For the LgA group, factor 2 from the PCA showed nega-
tive correlations with plasma levels of CCL-4 ( r = − 0.576, p = 0.019)
and TNF- 𝛼( r = − 0.556, p = 0.025) and a positive correlation with
plasma levels of IL-17 ( r = 0.611, p = 0.012; Fig. 6 B). We then corre-
lated the behavior comprised in factor 2 for the LgA group, hyperalge-
sia, and capsaicin-punished seeking, with IL-17, CCL-4 and TNF- 𝛼. Pun-
ished seeking showed an inverse relationship with CCL-4 ( r = − 0.658,
p = 0.006) and TNF- 𝛼( r = − 0.566, p = 0.022), and positive correlations
with hyperalgesia ( r = 0.476, p = 0.06) and IL-17 ( r = 0.586, p = 0.017).
4. Discussion
The present study found that (i) the number of fentanyl vapor deliv-
eries positively correlated with blood levels of fentanyl and its metabo-
lite, norfentanyl. (ii) Fentanyl vapor self-administration positively corre-
lated with antinociception. (iii) Mice that were given LgA to fentanyl es-
calated their intake, exhibited more signs of naloxone-precipitated with-
drawal, developed more robust hyperalgesia during spontaneous with-
drawal, and were more motivated to obtain fentanyl than mice given
ShA to fentanyl (iv) Fentanyl self-administration under the FR1 and PR
(both eort and time) schedules comprised the same behavioral con-
struct ( “intake/motivation ”) in both the ShA and LgA groups. In mice
in the LgA group, fentanyl self-administration despite punishment (i.e.,
punished taking) was associated with the drug “intake/motivation ”con-
struct. However, punished seeking and hyperalgesia in mice in the LgA
group comprised an independent construct ( “hyperalgesia/punished
seeking ”). Precipitated signs of withdrawal constituted a third, inde-
pendent factor in the LgA group. (v) Overall, plasma levels of corticos-
terone, chemokines, and cytokines were not dierent between the ShA
and LgA groups. However, in mice in the LgA group, more responding
during punished seeking positively correlated with IL-17 and negatively
correlated with CCL-4 and TNF- 𝛼.
We found that increasing the number of passive fentanyl vapor deliv-
eries resulted in higher blood levels of fentanyl and norfentanyl. With
7
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
Fig. 6. Correlation analysis . Individual
factor loadings from the PCA were corre-
lated with plasma levels of cytokines and
corticosterone in ShA (A) and LgA (B)
groups. Data were analyzed by Pearson’s
correlations using two-tailed p values with
𝛼= 1%. The p values are represented in-
side the boxes. Orange squares denote p <
0.05–0.001. Green squares denote p < 0.10.
( n = 7–8/sex/group).
this noncontingent drug delivery, no conditioned responses were ex-
pected to form, and the mice inhaled fentanyl vapor that dispersed
throughout the chamber. It is dicult to determine the extent to which
this drug concentration relates to drug use in humans, due to, for exam-
ple, inconsistent purity and cross-contamination with dierent drugs.
Kilmer et al. [37] reported an estimated use of fentanyl at 3–7 mg/day.
Applying allometric scaling [38] to the blood fentanyl levels in the mice,
the estimated concentrations would range from 0.37 to 2.33 ng/mL,
which is within the concentration used to induce analgesia in humans
[39] . Although we did not measure drug levels following operant self-
administration, we would argue that fentanyl levels are higher during
self-administration because the mice likely improve their navigation in
the operant chambers and quickly move close to the vapor delivery port
for greater inhalation [20] . Fentanyl vapor self-administration caused
analgesia, and analgesia positively correlated with the number of fen-
tanyl vapor deliveries. These ndings replicate previous studies with
vapor-delivered opioids [ 20 , 40 , 41 ] and validate our experimental con-
ditions, indicating that mice achieved sucient fentanyl levels during
self-administration that caused behavioral eects. We did not investi-
gate the development of tolerance to antinociceptive responses because
the LgA group escalated their intake in the rst hour of the session which
would be a confounding factor for the analysis.
We [ 18 , 19 , 20 , 35 ] and others [40–42] have developed opioid vapor
self-administration models in rodents. Mice and rats that are allowed
extended access to opioids escalate their intake over time and exhibit
greater motivation [ 13 , 43 , 14 ]. In our previous work, LgA mice were
given 12 h access to fentanyl [20] . Here, in male and female mice, 6 h
access to vaporized fentanyl was sucient to produce an escalation of
intake over time, and these mice exhibited an increase in dependence as
dened by naloxone-precipitated opioid withdrawal. We did not observe
major sex dierences in opioid intake or the rate of escalation ( Sup-
plemental Tables S1, S2 ), which contrasts with ndings in mice that
self-administered heroin intravenously or oxycodone orally, in which fe-
males had higher overall intake [ 15 , 44 ]. One potential explanation for
this dierence is drug-specic sex dierences (e.g., [45] ).
Both somatic and motivational withdrawal symptoms are caused by
the prolonged use of opioids. The somatic symptoms are generally ob-
served only during acute withdrawal, and are short lasting [46] . In ro-
dent models, both somatic and motivational signs of withdrawal can
be precipitated by the administration of µ-opioid receptor antagonists,
such as naloxone. Consistent with previous reports [ 14 , 18 , 15 , 20 ], we
found that mice in the LgA group exhibited more signs of precipitated
withdrawal compared with the ShA group ( Fig. 3 A).
The motivational component of addiction-like behaviors (e.g., in-
creases in drug taking and seeking) can be modeled in rodents with
such tasks as progressive-ratio tests [ 47 , 45 ]. We chose two tasks to as-
sess motivation: progressive-ratio schedules [47] in which the number
of operant responses (i.e., lever presses) and the time delay between an
operant response and reinforcer delivery that were necessary to obtain
the next reinforcer increased progressively. Mice in the LgA group ex-
hibited greater motivation to work ( Fig. 3 C) and wait ( Fig. 3 D) to obtain
the subsequent reinforcer compared with mice in the ShA group.
Another element that contributes to motivational drive in opioid de-
pendence is hyperalgesia (i.e., increase in pain sensitivity/decrease in
pain tolerance) during spontaneous withdrawal, an eect that is long-
lasting and hypothesized to contribute to addiction from the perspective
of drug seeking to relieve a negative emotional state [ 48 , 46 ]. Opioid
withdrawal-induced hyperalgesia is observed in humans [ 49 , 4 , 5 ] and
rodents [50–54] . Here, we observed that both ShA and LgA groups ex-
hibited a reduction of mechanical sensitivity thresholds compared with
baseline (i.e., before any opioid exposure), conrming that even low
doses of opioids can cause hyperalgesia, likely through a sensitization
process [50] , yet hyperalgesia was more pronounced in the LgA group.
Another construct that is associated with addiction-like behavior in
rodent models is drug seeking and taking despite adverse consequences
[55–57] . Commonly used punishers with opioid self-administration in-
clude taste adulterants, such as quinine, for oral self-administration
[58] ; irritants, such as histamine, for intravenous self-administration
[59] ; and mechanical punishers, such as air pu [60] and foot shock
[ 61 , 59 ]. Here, we used the respiratory irritant capsaicin as a punisher
for vaporized drug seeking and taking. Capsaicin activates nonselective
cation transient receptor potential vanilloid type 1 (TRPV1) channels
[62] . Heat and protons can activate TRPV1 channels on their own. Fur-
thermore, TRPV1 channels are upregulated in the sciatic nerve, dorsal
root ganglia, spinal cord [63] , sensory cortex, and thalamus [64] follow-
ing chronic morphine treatment. The administration of TRPV1 channel
antagonists in the nucleus accumbens signicantly reduces morphine
self-administration in rats [65] . In the hyperalgesic state, there is a
higher extracellular proton concentration and because protons can ac-
tivate TRPV1, this state potentiates responses to capsaicin, [62] , which
may have contributed to the apparent higher sensitivity of LgA mice to
punished fentanyl taking (see 0.3% on Fig. 4 B).
8
R.C.N. Marchette, E.R. Carlson, N. Said et al. Addiction Neuroscience 5 (2023) 100057
In the punished seeking test, capsaicin at dierent concentrations
was mixed with vapor vehicle (VG/PG), with no fentanyl in the solu-
tion. Thus, responding during this test was motivated by conditioned
positive and negative reinforcement responses [66] . Both groups equally
reduced their drug seeking, measured by vapor deliveries, when the ve-
hicle was adulterated with 0.03% and 0.1% capsaicin ( Supplemental
Table S3 ). To model punished taking, we adulterated the fentanyl solu-
tion with increasing concentrations of capsaicin so that the motivation
to lever press for drug in this test results from both conditioned eects
and drug eects. Both groups exhibited a reduction of drug taking at the
three highest capsaicin concentrations (0.3%, 1%, and 3%). The strong
analgesia and consequent hyperalgesic eects of opioids make it noto-
riously dicult to employ punishment in opioid self-administration. As
such, we found fewer studies in PubMed that employed punishment for
opioids (28 studies) vs. cocaine (69 studies) and alcohol (66 studies).
We conducted a PCA of data from the ShA and LgA groups sepa-
rately to gain a better understanding of interactions among multiple
dimensions that are associated with these two drug exposure condi-
tions. Quite similar patterns of behavioral constructs emerged for the
two groups. Escalation, re-escalation, and motivation (both eort and
delay) loaded onto one factor ( “intake/motivation ”). Punished taking
also loaded on this factor in the LgA group. Hyperalgesia and punished
seeking loaded on an independent factor ( “hyperalgesia/punished seek-
ing ”). Punished taking and naloxone-precipitated withdrawal loaded on
this factor in the ShA group. Although punished seeking was not higher
in the LgA group compared with the ShA group ( Fig. 4 , Supplemental
Table S4 ), the data suggest that higher hyperalgesia is associated with
higher aversion-resistant drug seeking in both groups. Heightened me-
chanical sensitivity was associated with higher opioid preference in mice
with a history of stress but not in controls [67] . Naloxone-precipitated
withdrawal in the LgA group loaded on a third independent factor, sug-
gesting dierent brain circuitries compared with hyperalgesia in LgA
mice [68] . These ndings highlight qualitative dierences between be-
havioral phenotypes of mice with dierent levels of drug access. In an-
imals with ShA to fentanyl, sensitivity to punishment, hyperalgesia and
somatic signs of dependence comprised the same behavioral construct,
and these variables may engage similar biological mechanisms. In