Extended access to fentanyl vapor self-administration leads to addiction-like behaviors in mice: Blood chemokine/cytokine levels as potential biomarkers

Abstract

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. Compared with mice in the ShA group, mice in the LgA group escalated their fentanyl intake, were more motivated 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 behavioral constructs: “intake/motivation” and “hyperalgesia/punished seeking.” In mice in the LgA condition only, “hyperalgesia/punished seeking” was associated with plasma levels of proinflammatory interleukin-17 (IL-17), chemokine (C-C motif) ligand 4 (CCL-4), and tumor necrosis factor α (TNF-α). Overall, the results suggest that extended 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 proinflammatory 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 clinically significant 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 opioids is a major problem in North America but has remained stable since 2010. Opioid overdose deaths doubled in the same period, suggesting an increase in the nonmedical use of potent opioids, 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 ).

Opioid use disorder involves a compulsion to seek and take opioids, the loss of control over intake, and the emergence of a negative affect state during withdrawal, also known as hyperkatifeia, which are symptoms that reflect 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 antagonist naloxone (Narcan^®) is used to reverse opioid overdose. Treatment 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 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, exhibit greater signs of naloxone-induced withdrawal, and are more motivated 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 vaporized 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. Additionally, most research has focused on reward and consummatory behaviors 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 fentanyl 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, elements 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 conditions, and (iii) drug seeking and taking despite aversive consequences [24].

Neuroimmune alterations in the brain have been shown to significantly alter key neurotransmitters involved in substance use disorders [25]. As such, the overproduction of cytokines and other immune signaling molecules can decrease dopamine and increase glutamate signaling and release (for review, see [26–28]). Conversely, prolonged exposure to an addictive drug affects the neuroimmune system by eliciting microglia and astrocyte activation, leading to the release of cytokines and chemokines [29]. Opioids induce many neuroinflammatory alterations in the brain that may contribute to the development of addiction through myriad mechanisms [29]. Opioids activate microglia (i.e., resident immune cells in the brain) and astrocytes (i.e., neuronal regulatory cells; [30]) leading to the release of proinflammatory cytokines [31]. Recent studies reported an upregulated expression of cytokines such as TNF-α, IL-1β, 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 cytokines 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 specific pro-inflammatory cytokines (TNF-α, IL-1β, 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 taking 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 cycle (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 Institutes 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 pharmacy. 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 experiments [20]. Capsaicin-adulterated fentanyl was used for drug seeking and taking despite punishment. Capsaicin (catalog no. N735–5 g, AK Scientific, Union City, CA, USA) was dissolved in 0.5 mL of 200-proof ethanol (catalog no. 111,000,200, Pharmco-Aaper, Brookfield, 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. 1A). 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 isoflurane 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).

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 cutoff) immediately before and after the fifth 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).

2.3. Nociception tests

To measure antinociception, we used a hot plate (Ugo Basile, Gemonio, Italy) at 52.5 °C. The latency to the first nociceptive behavior 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 flicking. A cutoff time of 30 s was imposed 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 removal). Two measures were taken for each rear paw with an interstimulus interval of at least 30 s. We calculated the average of these measures and used this as a measure of mechanical sensitivity. For principal component 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 environment. The pump produces negative pressure to pull fentanyl vapor inside the chambers. The outlet tubes connect to a HEPA filter before the vapor is released to the building exhaust system. All procedures were performed as previously described [35]. Briefly, mice were trained to lever press for vaporized fentanyl (1.5 s, 60 W, 1 min timeout) on a fixed-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 delivery was accompanied by activation of a cue light. The cue light remained 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 schedules (described below), nociception, and capsaicin-punished drug seeking. After 3 weeks of abstinence, all mice underwent 10 re-escalation sessions, after which they were tested for naloxone-precipitated withdrawal 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 delivery or a total of 6 h ended the session. The breakpoint was determined as the last ratio (“effort”) that was completed in the session.

2.6. Delay

We also tested the mice in a delayed-reward task, in which the interval between lever pressing and vapor delivery was sequentially increased 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 analyzed 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 increasing 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.

2.9. Plasma collection

After the end of the behavioral test battery, all mice underwent one final FR1 self-administration session (1 h for ShA and 6 h for LgA). After 36 h, the mice were deeply anesthetized with isoflurane and decapitated. 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 buffer, and the enzyme-linked immunosorbent assay (ELISA) was performed according to the manufacturer’s instructions (catalog no. ab108821, Abcam, Cambridge, UK).

2.11. Cytokine and chemokine measurement

Plasma samples were diluted by a factor of 1.5 and analyzed using an Ella microfluid 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 interactions and finding none (see below), we reanalyzed the male and female data combined using two-way repeated-measures ANOVA, followed 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 Supplemental 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 significant. To assess relationships among variables, we used Varimax-normalized principal components analysis (PCA) with Eigenvalues greater than 1. Loading factors ≥ 0.6 were considered. 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 positively correlated with each other (positive values) or negatively correlated with each other (negative values). We then extracted the individual 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 because of technical issues during plasma sample processing.

3. Results

3.1. Sex differences

We analyzed all data for sex effects (Supplemental Tables S1–4). We employed three-way ANOVAs and found no sex × group interactions for any of the variables analyzed. An overall effect of sex on punished seeking 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). Because 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 confirmed 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. 1B). 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. 1C). 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 effect (i.e., antinociception).

3.3. Escalation of fentanyl vapor self-administration in dependent mice

For the first hour of the session, the two-way ANOVA showed a significant group × session interaction (F _9,270 = 3.880, p = 0.00012; Fig. 2A). 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 significant group × session interaction for the number of vapor deliveries (F_9,270 = 2.917, p = 0.0026; Fig. 2B). 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 differences in vapor deliveries across sessions in the ShA group (Fig. 2A, B). Using linear regression, we calculated the slope of the escalation curve for each mouse. Mice in the LgA group had significantly higher slopes than mice in the ShA group, further confirming their escalation of intake (t ₃₀ = 2.259, p = 0.03; Fig. 2C). For the first hour of re-escalation, the two-way repeated-measures ANOVA did not show a significant group × session interaction (F _9,270 = 1.880, p = 0.055; Fig. 2D) nor main effect of group (F_1,30 = 1.657, p = 0.21) but showed a main effect 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 significant group × session interaction for the number of vapor deliveries for the total 6 h during re-escalation (F_9,270 = 1.567, p = 0.12; Fig. 2E) but showed main effects 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 overall 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 differ between ShA and LgA conditions (t₃₀ = 0.48, p = 0.63; Fig. 2F). Mice allowed LgA escalated their fentanyl intake during the escalation phase and maintained higher levels of fentanyl intake but did not escalate further on the re-escalation phase.

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 first 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 first 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 differ 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).

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 exhibited more signs of naloxone-precipitated withdrawal compared with mice in the ShA group (t₃₀ = 3.424, p = 0.002; Fig. 3A). Unpaired Student’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 hyperalgesia (t₃₀ = 2.409, p = 0.049; Fig. 3B). 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 (t₃₀ = 2.299, p = 0.03; Fig. 3C). 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 (t₃₀ = 2.090, p = 0.04; Fig. 3D). In summary, mice allowed LgA to fentanyl self-administration showed more signs of somatic withdrawal, more hyperalgesia and were more motivated to obtain fentanyl than the ShA group.

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”effort”). 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).

3.5. Self-administration despite punishment

For punished fentanyl seeking, the two-way repeated-measures ANOVA showed no difference between groups (p > 0.05) and no group × capsaicin concentration interaction (p > 0.05) but a significant effect of capsaicin concentration (F_3,90 = 12.60, p < 0.0001; Fig. 4A), 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 significant group × capsaicin concentration interaction (F_6,180 = 3.797, p = 0.0014; Fig. 4B). The post hoc comparisons did not indicate a difference 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% capsaicin (p = 0.001) was significantly 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 significantly 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.

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).

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 behavioral variables, combining males and females. We used the average of the last three escalation sessions (i.e., sessions that were different 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-specified 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 associations between fentanyl intake (escalation and re-escalation) and motivation (PR and delay). Factor 2 comprised 22.73% of the data variance and included positive associations between hyperalgesia, capsaicin-punished seeking and taking, and naloxone-precipitated withdrawal. For the LgA group, factor 1 (42.42% of the variance) showed positive associations between fentanyl intake (escalation and re-escalation), motivation (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) comprised naloxone-precipitated withdrawal signs. These data suggest that there are qualitative differences on the behavioral phenotype of mice allowed ShA or LgA to fentanyl. Although PR and delay measures (motivation) 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.

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).

3.7. Plasma levels of cytokines and corticosterone

There were no significant group differences in plasma corticosterone levels (p > 0.05), or cytokine and chemokine levels (p > 0.05), including iIL-1β, IL-6, IL-17, CCL-2, CCL-4, and TNF-α (Table 1).

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-1β (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

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).

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 Supplemental Table S5). We did not observe any correlations for the ShA group (Fig. 6A). For the LgA group, factor 2 from the PCA showed negative 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. 6B). We then correlated the behavior comprised in factor 2 for the LgA group, hyperalgesia, and capsaicin-punished seeking, with IL-17, CCL-4 and TNF-α. Punished 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).

Fig. 6. Correlation analysis.

Individual factor loadings from the PCA were correlated 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 inside the boxes. Orange squares denote p < 0.05–0.001. Green squares denote p < 0.10. (n = 7–8/sex/group).

4. Discussion

The present study found that (i) the number of fentanyl vapor deliveries positively correlated with blood levels of fentanyl and its metabolite, norfentanyl. (ii) Fentanyl vapor self-administration positively correlated with antinociception. (iii) Mice that were given LgA to fentanyl escalated their intake, exhibited more signs of naloxone-precipitated withdrawal, developed more robust hyperalgesia during spontaneous withdrawal, and were more motivated to obtain fentanyl than mice given ShA to fentanyl (iv) Fentanyl self-administration under the FR1 and PR (both effort and time) schedules comprised the same behavioral construct (“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” construct. 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, independent factor in the LgA group. (v) Overall, plasma levels of corticosterone, chemokines, and cytokines were not different 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 deliveries resulted in higher blood levels of fentanyl and norfentanyl. With this noncontingent drug delivery, no conditioned responses were expected to form, and the mice inhaled fentanyl vapor that dispersed throughout the chamber. It is difficult to determine the extent to which this drug concentration relates to drug use in humans, due to, for example, inconsistent purity and cross-contamination with different 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 fentanyl vapor deliveries. These findings replicate previous studies with vapor-delivered opioids [20,40,41] and validate our experimental conditions, indicating that mice achieved sufficient fentanyl levels during self-administration that caused behavioral effects. We did not investigate the development of tolerance to antinociceptive responses because the LgA group escalated their intake in the first 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 sufficient to produce an escalation of intake over time, and these mice exhibited an increase in dependence as defined by naloxone-precipitated opioid withdrawal. We did not observe major sex differences in opioid intake or the rate of escalation (Supplemental Tables S1, S2), which contrasts with findings in mice that self-administered heroin intravenously or oxycodone orally, in which females had higher overall intake [15,44]. One potential explanation for this difference is drug-specific sex differences (e.g., [45]).

Both somatic and motivational withdrawal symptoms are caused by the prolonged use of opioids. The somatic symptoms are generally observed only during acute withdrawal, and are short lasting [46]. In rodent 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. 3A).

The motivational component of addiction-like behaviors (e.g., increases in drug taking and seeking) can be modeled in rodents with such tasks as progressive-ratio tests [47,45]. We chose two tasks to assess 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 exhibited greater motivation to work (Fig. 3C) and wait (Fig. 3D) to obtain the subsequent reinforcer compared with mice in the ShA group.

Another element that contributes to motivational drive in opioid dependence is hyperalgesia (i.e., increase in pain sensitivity/decrease in pain tolerance) during spontaneous withdrawal, an effect that is longlasting 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 exhibited a reduction of mechanical sensitivity thresholds compared with baseline (i.e., before any opioid exposure), confirming 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 include 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 puff [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. Furthermore, TRPV1 channels are upregulated in the sciatic nerve, dorsal root ganglia, spinal cord [63], sensory cortex, and thalamus [64] following chronic morphine treatment. The administration of TRPV1 channel antagonists in the nucleus accumbens significantly reduces morphine self-administration in rats [65]. In the hyperalgesic state, there is a higher extracellular proton concentration and because protons can activate 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. 4B).

In the punished seeking test, capsaicin at different concentrations was mixed with vapor vehicle (VG/PG), with no fentanyl in the solution. 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 vehicle was adulterated with 0.03% and 0.1% capsaicin (Supplemental Table S3). To model punished taking, we adulterated the fentanyl solution with increasing concentrations of capsaicin so that the motivation to lever press for drug in this test results from both conditioned effects and drug effects. 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 effects of opioids make it notoriously difficult 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 separately to gain a better understanding of interactions among multiple dimensions that are associated with these two drug exposure conditions. Quite similar patterns of behavioral constructs emerged for the two groups. Escalation, re-escalation, and motivation (both effort 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 seeking”). 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 mechanical 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, suggesting different brain circuitries compared with hyperalgesia in LgA mice [68]. These findings highlight qualitative differences between behavioral phenotypes of mice with different levels of drug access. In animals 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 mice allowed LgA to fentanyl, a different behavioral phenotype was observed. Somatic signs of dependence were independent from all behavioral measures. Additionally, in mice in the LgA group, the mechanisms that underlie resistance to punished drug taking, motivation for fentanyl and fentanyl intake likely shared biological mechanisms as well as the mechanism underlying sensitivity to punished drug seeking and hyperalgesia. It is important to note that these are correlative results, and studies will be needed to test their functional relationship.

Given that PCA factors are considered independent from each other, the biological mechanisms that underlie these behavioral constructs are also likely different. Consistent with this hypothesis, punishment insensitivity may constitute a unique phenotype that is independent from reward seeking and Pavlovian fear [69] and separate from motivation for alternative rewards and the drug [70]. Aoun et al. found that”intake/motivation” and “punished drinking/anxiety” emerged as independent constructs in alcohol-dependent and nondependent rats [71]. Rats that were identified to be shock-resistant alcohol drinkers were also more resistant to quinine adulteration, but they were less motivated for alcohol in a progressive-ratio test [72]. These findings suggest that conflict is a component of “aversion resistance.” Aoun et al. [71] used the elevated plus-maze conflict model of anxiety; herein, we presented a conflict between receiving drug/drug-associated cue and punishment. The inhibition of synaptotagmin-2 in the prefrontal cortex increased aversion-resistant alcohol (adulterated with quinine) drinking without changing the FR1 self-administration of non-adulterated alcohol [73], providing support for different molecular mechanisms between intake and aversion-resistant intake. However, motivation, resistance to punishment, and drug seeking were correlated in cocaine self-administering rats [55]. Thus, LgA to drugs to the point of dependence (i.e., the manifestation of withdrawal) in preclinical models can increase various behaviors, such as escalation, motivation, hyperalgesia, and punished responding that comprise different behavioral constructs [13,74,18,20,54,75] that contribute to the overall phenotype of addiction-like behavior and provide insights into biological mechanisms.

Lastly, we searched for correlations between behavioral constructs and blood chemokine/cytokine levels. Overall, we did not observe differences between the ShA and LgA groups in plasma levels of corticosterone and chemokines/cytokines (IL-1β, IL-6, CCL-2, CCL-4, and TNF-α). We are unaware of studies that investigated blood chemokine/cytokine levels in opioid dependence. Mahajan et al. [34] reported downregulated mRNA expression of CCL2 and CCL4 chemokines in human astrocytes following chronic morphine exposure [34], whereas other studies revealed that prolonged administration of oxycodone and withdrawal raised protein levels of proinflammatory cytokines (TNF-α, IL-1β, IL-6, and IL-17) [32]. These findings corroborate other studies that associated opioid exposure with changes in neuroimmune markers [31,51,76–79] in rodent brain tissue. Opioids may also cause the reactivity of microglia and astrocytes in rodent brain tissue [30] and human brain tissue (e.g., central nucleus of the amygdala; [45]). In the ShA group, we did not find correlations between principal component factor 1 (escalation, re-escalation, PR, delay) and factor 2 (, hyperalgesia, precipitaded withdrawal, punished seeking and punished taking) with corticosterone or cytokines. However, in the LgA group, the “hyperalgesia/punished seeking” factor was associated with CCL-4, TNF-α, and IL-17, all of which are proinflammatory cytokines. This association was driven more by the punished seeking behavior than hyperalgesia. Further analysis confirmed that punished seeking positively correlated with IL-17 and negatively correlated with CCL-4 and TNF-α in the LgA group. These preliminary findings encourage further investigations of chemokines/cytokines as potential biomarkers and/or mediators of opioid dependence in blood and other tissues.

5. Conclusion

Our results suggest that extended fentanyl vapor self-administration leads to the development of addiction-like behaviors. The measurement of several opioid-related behaviors allowed us to identify independent behavioral constructs that may capture different aspects of opioid seeking with extended access. We also provided preliminary evidence of the potential of blood proinflammatory chemokine/cytokine levels to serve as biomarkers of opioid-related behaviors in opioid dependence.

Data availability

Data will be made available on request.