Acute and persistent sex-dependent disruptions of sleep-wake architecture induced by fentanyl self-administration in mice

Abstract

Chronic opioid use is associated with profound sleep disturbances that may contribute to relapse in opioid use disorder (OUD). However, how opioid self-administration and abstinence alter sleep architecture in a sex-dependent manner remains poorly understood. Here, we examined the effects of volitional fentanyl intravenous self-administration (IVSA) on sleep-wake architecture in male and female mice across drug-taking and abstinence phases. Continuous home-cage monitoring revealed sex differences in baseline sleep, with females exhibiting increased wakefulness and reduced NREM sleep relative to males across conditions. Fentanyl IVSA increased dark-phase NREM sleep following self-administration sessions, whereas abstinence was characterized by a persistent reduction in REM sleep accompanied by increased REM bout frequency, indicative of REM fragmentation. Correlational analyses between sleep architecture and IVSA behavior revealed that higher fentanyl intake was associated with reduced sleep, particularly in females, in contrast to the positive sleep-intake relationship observed in saline controls. Principal component analysis identified a primary sleep component that captured fentanyl-related alterations in sleep–behavior relationships and a secondary component reflecting sex-dependent variation in sleep architecture. Together, these findings demonstrate that fentanyl self-administration induces both acute and enduring, sex-specific disruptions in sleep architecture that persist into abstinence and are associated with drug-seeking behavior. These results highlight sleep as a potential mechanistic contributor to relapse vulnerability and a promising therapeutic target in OUD.

1. Introduction

Opioid use disorder (OUD) is an ongoing public health concern. The current wave of the opioid crisis is characterized using synthetic opioids, most notably fentanyl(Centers for Disease Control and Prevention 2021). Fentanyl is a synthetic opioid that is estimated to be 10 times more potent than heroin(Bird, Huhn et al. 2023). As such, within the past decade, fentanyl has been responsible for a surge in drug-related overdoses, and for most opioid-related overdoses(Ahmad, Cisewski et al. 2025). In addition to overdose, chronic opioid misuse is associated with significant sleep and circadian disruptions(Greenwald, Moses et al. 2021). Specifically, individuals with OUD have reported the most severe disruption of both non-rapid eye movement sleep (NREMS) and rapid eye movement sleep (REMS) occurs during withdrawal and within the first few weeks of abstinence(Lewis, Oswald et al. 1970, Kay, Pickworth et al. 1981). Given these sleep disturbances and the lack of medications for the treatment of OUD, which currently consist primarily of other opioids (e.g., buprenorphine, methadone, etc.), there remains a high risk of relapse, wherein between 59% and 80% of individuals will relapse within 1 week to 1 month of cessation of drug use, and as high as 90% will relapse during recovery(Smyth, Barry et al. 2010). Notably, studies have also suggested that a major cause of relapse in individuals who previously misused opioids are sleep disturbances(Brower and Perron 2010), highlighting the need to study sleep in the context of OUD.

Self-administration is considered the gold standard of animal models as it can recapitulate the drug-taking and seeking patterns of individuals with substance use disorders. Despite the knowledge that chronic opioid misuse can cause sleep disturbances, very few preclinical studies have examined this phenomenon. Indeed, our lab has previously shown that fentanyl can alter sleep architecture and diurnal pattern of sleep-wake states in male mice, specifically revealing that disruption of NREMS increases during withdrawal(Gamble, Chuan et al. 2022). However, volitional opioid use has yielded inconsistent results. For example, a recent study found that chronic sleep deprivation did not induce morphine preference in a 2-bottle choice paradigm in male rats but still produced conditioned place preference(Eacret, Lemchi et al. 2022). Furthermore, the effects of opioids have been shown to be sex-dependent in both clinical and preclinical studies. Indeed, women display higher susceptibility to opioid dependence, and the efficacy of opioids is different in women compared to men(Greenfield, Manwani et al. 2003, Niesters, Dahan et al. 2010, Back, Payne et al. 2011). Additionally, sex differences have been reported in sleep disturbances due to OUD(Finan, Mun et al. 2020, He, Tang et al. 2020). Supporting this, a recent study highlighted that female mice displayed a more pronounced withdrawal-associated sleep phenotype, indicating a greater sensitivity to morphine and/or faster escalation to dependence(Tisdale, Sun et al. 2024). Consistent with these observations, a different study showed that heroin self-administration resulted in reverse sleep-wake cycles in rats, specifically showing persistent REM disruption into early abstinence(Coffey, Guan et al. 2016). Overall, these studies emphasize the relationship between opioid use and sleep and warrants further investigation of the effects of synthetic opioids (i.e., fentanyl), given their current prevalence.

Here, we used a model of fentanyl self-administration combined with noninvasive sleep phenotyping to determine whether volitional fentanyl use and seeking results in sleep disruption. To our knowledge, this is the first study to examine the effect of fentanyl self-administration on sleep-wake states across different phases of the addiction cycle – drug-taking, abstinence, and cue-induced drug-seeking (a model of drug craving). Additionally, we further analyze our results to investigate whether specific self-administration behaviors correlate with disrupted sleep-wake states to find potential “risk traits” that may predict further drug use and relapse.

2. Material and Methods

2.1. Animals

Male and female C57BL/6J mice (8–12 weeks; Jackson Laboratories, Bar Harbor, ME; #000664) were maintained on a 12:12 reverse light–dark cycle (lights on at 7 PM, zeitgeber time (ZT) 0, and lights off at 7 AM, ZT12), at 22 ± 1.5 °C. Animals were provided with food and water ad libitum. All animals were maintained according to the National Institutes of Health guidelines and approved by the University of Massachusetts Chan Medical School Institutional Animal Care and Use Committee. Animal sample size was justified by previously published data or preliminary experiments.

2.2. Drug administration

Fentanyl powder was purchased from Supelco (PHR8976, Millipore Sigma) and dissolved in 0.9% Saline then filtered through 0.2 mm filter (431224, Corning). Mice were randomly assigned to groups and counterbalanced by sex. Self-administration sessions occurred between ZT15–17. A total of 28 mice were used: 12 saline (6 females, 6 males) and 16 fentanyl (8 females, 8 males) during the habituation and administration phases. Tissue from 8 mice during the intoxication phase was collected for a separate study immediately after the last session. Therefore, the abstinence phase included the remaining 20 mice (a subset of the original 28): 10 saline (5 females, 5 males) and 10 fentanyl (5 females, 5 males). All mice were placed into the self-administration chambers without prior training and spontaneously nose-poked for infusions; no shaping or operant training was required.

2.3. Jugular catheterization

Mice were anesthetized with isoflurane (1–3% mixed with oxygen). An indwelling catheter was passed subscapularly and implanted into the right jugular vein. The catheter is prepared using a 5.5 cm length of SILASTIC tubing (0.012” inner and 0.025” outer diameters) threaded onto a vascular access button (VABM1B/25, Instech Laboratories), which exits on the back of the mouse and is protected by an aluminum cap. Mice were administered a Meloxicam injection (4 mg/kg, s.c.) diluted in sterile 0.9% Saline prior to start of surgery. Catheters were flushed daily with sterile 0.9% Saline containing heparin (10–30 U/mL) beginning 48 hours after surgery. Animals were allowed at least 3 days to recover (average recovery time was 5 days) from surgery before the first self-administration session.

2.4. Sleep and circadian monitoring

Following jugular catheterization, mice underwent non-invasive three-state sleep-wake analysis using PiezoSleep Mouse Behavioral Tracking Systems (Signal Solutions, Inc., Lexington, KY, USA). Mice were singly housed and provided food and water ad libitum, under identical reverse light–dark cycle conditions within sound attenuating and light-controlled cabinets (Tecniplast; Westchester, PA, USA). Mice were allowed to habituate for 3 days prior to data acquisition and the first self-administration session, wherein they were briefly removed from the sleep chambers.

2.5. Self-administration

Mice were trained to self-administer Fentanyl or 0.9% Saline in 15, 2-h sessions (5 days on/2 days off) under a fixed ratio 1 (FR1) schedule of reinforcement in standard operant chambers (Med Associates). Successful completion of criteria resulted in delivery of a Fentanyl (3 μg/kg/inf) or 0.9% Saline infusion and activation of a cue-light and 2.5 kHz tone, as well as a 20-sec timeout period in which active nose pokes were recorded but carried no scheduled consequences. Inactive nose pokes were also recorded but carry no scheduled consequences. Once stable responding was achieved (>10 infusions, 2:1 ratio of active vs inactive nose pokes, and <25% variation across 3 consecutive sessions), mice underwent 4 weeks of abstinence. Cue-induced drug-seeking was assessed on D1 and D28 of abstinence under original self-administration conditions; however, no infusions were delivered. Animals that did not display stable responding were excluded from analyses. In the current study, no mice were excluded from analysis due to lack of stable responding.

2.6. Statistical analyses

2.6.1. Behavior

Data were analyzed by Two-way RM ANOVA. Data from cue-induced drug-seeking sessions were analyzed using Two-way ANOVA. Significant results were followed by appropriate post hoc tests (Tukey’s or Fisher’s LSD). Averaged summary data were assessed for normality using a Shapiro-Wilk test, then analyzed as t-tests (Mann-Whitney U for nonparametric measures) with Welch’s correction applied when F test to compare variances failed. Data are presented as mean ± standard error of mean (SEM) with significance set at α = 0.05. Data were analyzed using Microsoft Excel and GraphPad Prism software (version 10.6.1).

2.6.2. IVSA behavioral measures

IVSA data were extracted from session timestamp files using our laboratory’s custom R pipeline. For each session and subject, the following variables were computed: active and inactive nose poke responses, total responses, number of infusions, and percent accuracy. Latency to the first active response, inter-infusion interval, and inter-active response interval were calculated to assess response timing. Within-session escalation was quantified as the slope of infusion counts across successive 5-minute bins using linear regression. Timeout responses were defined as active responses occurring within 20 seconds of an infusion. Response rates were calculated per 5-, 15-, and 30-minute intervals. Weekly escalation slopes were also computed across the five sessions per week.

2.6.3. Wake, NREM, and REM assessment via PiezoSleep and bout analysis

Sleep-wake behavior was continuously recorded in home cages using the PiezoSleep system (Signal Solutions, Inc., Lexington, KY, USA). Raw signals were classified into wake, NREM, and REM states using SleepStats software (Signal Solutions, Inc., Lexington, KY, USA). Data were exported as .awd files, which were imported into ClockLab (v6.0, ActiMetrics, Lafayette, IN) for bout analyses. Bout metrics included number, and average counts per bout. In addition, total time and percentage of time spent in each sleep state were calculated. All data were aggregated across baseline, drug-taking, and abstinence periods.

2.6.4. Z-score computation for IVSA and sleep behavioral measures

To standardized behavioral metrics, z-scores were calculated for each IVSA variable across weeks 1–3 of drug taking, and for sleep variables during habituation, weeks 1–3 of drug taking, and early (weeks 1–2) and late (weeks 3–4) abstinence. Variables where higher values indicate reduced motivation or slower responding (latency, inter-infusion interval, inter-active response interval) were multiplied by −1 prior to standardization. Z-scores were computed across all subjects for each variable using the formula: $z=\frac{x-\mu }{\sigma }$. Standardized IVSA and sleep variables are provided in Supplementary Table 1 and Supplementary Table 2 respectively.

2.6.5. Principal component analysis (PCA) and principal component (PC) loadings of sleep and IVSA measures

Following the computation of z-scores for all behavioral variables, PCA was performed separately for sleep- and IVSA-related measures for each subject using prcomp() in R. The first two principal components for sleep (Sleep PC1, Sleep PC2) and IVSA (IVSA PC1, IVSA PC2) were extracted for each subject. PCA scores for all subjects are provided in Supplementary Table 3. Relationships between corresponding PCs (Sleep PC1 vs. IVSA PC1; Sleep PC2 vs. IVSA PC2) were evaluated within each sex (female, male) and group (saline, fentanyl) using linear regression PC = β₀ + β₁ · Sleep_PC + ε. To determine the contribution of individual variables to overall variance, absolute PC1 and PC2 loadings were extracted for each variable (Lintz, Liu et al. 2025) Supplementary Table 4.

2.6.6. Correlation analysis between sleep architecture and IVSA behavior

Pearson correlation coefficients were calculated to examine the relationship between sleep and IVSA parameters. Analyses were conducted separately for females and males within saline and fentanyl groups for each IVSA week (Weeks 1–3) and for two abstinence epochs (early and late). Two correlation analyses were performed: (1) correlations between all individual sleep variables and IVSA PC1, and (2) correlations between Sleep PC1 and individual IVSA variable. Pearson correlation coefficients (r) and two-tailed p-values were computed in R. Corresponding correlation values and significance reported in Supplementary Tables 5 and 6.

2.6.7. Linear modeling of sleep variables versus IVSA principal components

To assess how IVSA behavior relates to sleep, linear regression models were fit for each sleep variable using the first two IVSA principal components (IVSA PC1 and IVSA PC2) as predictors. Each model included group (saline, fentanyl) and sex (female, male) as covariates to account for group and sex effects. Linear models were constructed using the lm() function in R. For each model, the coefficient estimates (β), standard errors, and R² values were recorded. P-values were corrected for multiple comparisons using the false discovery rate (FDR) method. Results for IVSA PCs, sex, and group are provided in Supplementary Table 7.

3. Results

3.1. Sex- and group-specific patterns of operant responding and fentanyl intake

An overview of the experimental design is shown in Fig. 1A. Both females and males reliably discriminated between active and inactive nose ports across self-administration sessions (3-way ANOVA session × group × port; main effect of port: females F(1,24) = 316.6, p < 0.0001; males F(1,24) = 123.2, p < 0.0001) (Fig. 1B–C). In females, responding varied by group over sessions (session × group: F(4.38,105.0) = 2.47, p = 0.044) (Fig. 1B), with no significant saline versus fentanyl differences at individual sessions. Males showed no session- or group-related effects (Fig. 1C). Overall, females earned more fentanyl infusions than males across sessions (2-way ANOVA sex × session: main effect of sex: F(1,14) = 13.35, p = 0.0026, session: F(5.3, 74.2) = 3.11, p = 0.0118) (Fig. 1D).

Figure 1. Experimental timeline and self-administration behavioral outcomes.

(A) Experimental timeline. Mice (n=5–8 per sex/treatment) underwent fentanyl (3 μg/kg/inf) or saline self-administration. Mice were habituated to PiezoSleep chambers for 3 days prior to the first self-administration session, followed by 4 weeks of abstinence. (B) Active and inactive nose poke responses for females, and (C) males. (D) Infusions earned across sessions. (E) Cue-induced seeking session were performed on abstinence day 1 and day 28. *p < 0.05, **p < 0.01

Females showed session-dependent changes in intake (2-way ANOVA group × session: main effect of session: F(4.66,55.91) = 5.02, p = 0.0009) with no group effect or interaction (Supplementary Fig. 1A), while males maintained stable intake across sessions regardless of group (Supplementary Fig. 1B). In the fentanyl groups, active port responding differed by sex, with females responding more than males across sessions (2-way ANOVA sex × session: main effect of sex: F(1,14) = 10.42, p = 0.0061; Supplementary Fig. 1C), while no sex differences were observed in the saline groups (Supplementary Fig. 1D).

During early (day 1) and late (day 28) abstinence, changes in responding differed between active and inactive ports (3-way ANOVA day × group × port: day × port interaction: F(1,36) = 14.26, p = 0.0006), with post hoc analyses showing that active responding decreased in saline mice (Tukey’s, p = 0.0001) but remained stable in fentanyl mice (Tukey’s, p = 0.17), indicating persistent operant responding in abstinence following fentanyl but not saline self-administration (Fig. 1E).

3.2. Fentanyl increased dark phase NREM sleep following IVSA sessions

To assess the effects of fentanyl on sleep, NREM, REM, and wake patterns were measured in females and males following fentanyl or saline IVSA. Dark-phase NREM was significantly increased after fentanyl IVSA compared to saline (3-way ANOVA group × sex × week, effect of group: F(1,24) = 4.33, p = 0.0484), whereas no group differences were observed at baseline or during early or late abstinence (p > 0.05; Fig. 2A). Averaging across weeks 1–3 confirmed this effect (2 way ANOVA group × sex, main effect of group: F(1,24) = 4.26, p = 0.0499; Fig. 2B). Dark-phase wake trended lower in fentanyl animals (main effect of group: F(1,24) = 3.906, p = 0.0597; Fig. 2C).

Figure 2. Fentanyl IVSA increases dark-phase NREM post-session, trends toward reduced wake time.

(A) Weekly percent NREM time in the dark phase across baseline, three weeks of fentanyl or saline intravenous self-administration (IVSA), and four weeks of abstinence (homecage, no drug) in females and males. (B) Percent of time spent in NREM and (C) wake during the dark phase after saline or fentanyl IVSA. (D) Number of NREM bouts and (E) wake bouts in the dark phase after IVSA sessions. Data are mean ± SEM; sample sizes: saline (n=6/sex), fentanyl (n=8/sex). Three-way repeated-measures ANOVA was used for weekly comparisons, and two-way ANOVA for total percentages and bout counts.

Analysis of sleep bout structure revealed sex-specific effects. Dark-phase NREM bout number was higher in males overall (main effect of sex: F (1, 24) = 8.645, p = 0.0071) (Fig. 2D). Dark-phase wake bout number differed by sex, with females showing fewer bouts than males (main effect of sex: F(1,24) = 13.09, p = 0.0014). Post hoc testing indicated that this difference was significant specifically in the saline group (Tukey’s p = 0.0289; Fig. 2E). Together, these results indicate that fentanyl IVSA increases dark-phase NREM and subtly reduces wake, with additional sex-specific differences in NREM and wake bout patterns.

3.3. REM decreased during abstinence following fentanyl IVSA

To assess how fentanyl affects sleep during abstinence, NREM, REM, and wake patterns were measured in females and males following 3 weeks of fentanyl or saline IVSA. Across the 24-hour day, mice that underwent fentanyl IVSA exhibited reduced REM sleep in abstinence compared to saline controls (2-way ANOVA sex × group: main effect of group: F(1,16) = 7.45, p = 0.0149) (Fig. 3A–B). This effect was driven by decreases in light-phase REM (main effect of group: F(1,16) = 10.49, p = 0.0051) with no significant changes in the dark phase (Fig. 3C). Fentanyl-exposed mice showed higher average counts per bout over 24 hours (main effect of group: F(1,24) = 11.03, p = 0.0029), but post hoc comparisons between males did not reach significance (Tukey’s p = 0.0560; Fig. 3D). Total bout number over 24 hours showed a trend toward an effect of group but did not reach statistical significance (F(1,24) = 4.096, p = 0.0543; Fig. 3E). Overall, following fentanyl IVSA, abstinence is characterized by reduced total and light-phase REM sleep.

Figure 3. REM decreased during abstinence following fentanyl IVSA, with increased bout number and average counts per bout.

(A) Percent of time spent in REM over 24 hours during 4 weeks of abstinence in males and females previously allowed to self-administer fentanyl or saline. (B) Percent of REM over the full 24-hour day, and (C) during light phase. (D) Average counts per bouts and (E) number of activity bouts during abstinence. Data are mean ± SEM; sample sizes: saline (n=5/sex), fentanyl (n=5/sex). Two-way ANOVA for total percentages and bout counts.

3.4. Sex Differences in Wake and NREM Across Baseline, IVSA, and Abstinence

At baseline, females exhibited more wake and less NREM than males across 24h and the light phase (two-way ANOVA: sex × group: wake total F(1,24) = 8.64, p = 0.0072; wake light F = 5.84, p = 0.0236; NREM total F(1,24) = 12.85, p = 0.0015; NREM light F = 8.22, p = 0.0085), with no sex differences in REM or dark-phase sleep (p > 0.05) (Supplementary Fig. 2A–I).

These differences persisted during IVSA, driven by the light phase (NREM total F(1,24) = 19.45, p = 0.0002; NREM light F(1,24) = 30.44, p < 0.0001; wake total F(1,24) = 8.360, p = 0.0080; wake light F(1,24) = 23.46, p < 0.0001; Supplementary Fig. 2A–D), with no sex differences in REM (p > 0.05) (Supplementary Fig. 2E–G).

During abstinence, females progressively showed less NREM and more wake than males across 24 hours. In week 1, sex differences were not significant for either NREM (total: F(1,16) = 3.39, p = 0.084; light: F(1,16) = 2.82, p = 0.112; dark: F(1,16) = 0.64, p = 0.436) or wake (total: F(1,16) = 2.86, p = 0.110; light: F(1,16) = 2.29, p = 0.150; dark: F(1,16) = 0.55, p = 0.471). By week 2, females exhibited significantly reduced NREM (total: F(1,16) = 19.85, p = 0.0004; light: F(1,16) = 10.31, p = 0.0054; dark: F(1,16) = 6.92, p = 0.018) and increased wake (total: F(1,16) = 13.47, p = 0.0021; light: F(1,16) = 4.59, p = 0.048; dark: F(1,16) = 7.94, p = 0.012). These sex differences intensified in week 3 for NREM (total: F(1,16) = 56.57, p < 0.0001; light: F(1,16) = 19.08, p = 0.0005; dark: F(1,16) = 16.45, p = 0.0009) and wake (total: F(1,16) = 45.22, p < 0.0001; light: F(1,16) = 10.05, p = 0.0059; dark: F(1,16) = 15.89, p = 0.0011), and remained significant in week 4 (NREM total: F(1,16) = 48.28, p < 0.0001; light: F(1,16) = 14.21, p = 0.0017; dark: F(1,16) = 8.49, p = 0.010; wake total: F(1,16) = 37.84, p < 0.0001; light: F(1,16) = 8.67, p = 0.0095; dark: F(1,16) = 9.25, p = 0.0078; Supplementary Figs. 4–5). Together, these results indicate that sex differences in NREM and wakefulness emerged after the first week of abstinence and persisted throughout later weeks. Overall, females consistently showed more wake and less NREM than males across baseline, IVSA, and abstinence, whereas REM did not differ by sex.

3.5. Sleep-behavior correlations vary by drug and sex

Correlations between sleep measures and IVSA behavior were generally modest across sex, group, and phase of addiction. Female fentanyl mice in week 1 of IVSA showed the strongest relationship for REM sleep (REM Total: r = 0.92, p = 0.52), higher than any other group (Fig. 4A). During abstinence, total 24h REM was correlated with IVSA behavior in female fentanyl mice in early abstinence (r = 0.89, p = 0.93) and late abstinence (r = 1.00, p = 0.46); while correlations were strong, p values did not reach significance, reflecting the reduced sample size in abstinence. Male fentanyl mice showed moderate, nonsignificant correlations in early abstinence (r = 0.59, p = 0.93) and late abstinence (r = 0.77, p = 0.93), whereas saline groups showed weaker, nonsignificant correlations (r = −0.44–0.81, p > 0.05) (Fig. 4A; Supplementary Table 5). Other sleep-IVSA behavior correlations, including NREM and wake, were generally smaller and nonsignificant, highlighting REM as the primary sleep measure associated with behavior in fentanyl-exposed animals.

Figure 4. Fentanyl IVSA Impacts REM Sleep in a Sex- and Phase-Specific Manner and Shifts Sleep-IVSA Correlations Opposite to Saline.

(A) Heatmap showing Pearson correlations between IVSA principal component 1 (PC1) and multiple sleep measures (Wake, NREM, REM) across three weeks of IVSA and four weeks of abstinence. Columns represent sex (female, male) and treatment (saline, fentanyl) groups across weekly IVSA sessions (Week 1–3) and abstinence periods (early = weeks 1–2, late = weeks 3–4). (B) Heatmap showing Pearson correlations between IVSA behavioral variables and the first principal component of sleep (Sleep PC1) across the same timepoints. Columns represent IVSA variables, rows represent timepoints, with females in the top rows and males in the bottom rows. Positive correlations are shown in red, negative correlations in blue, and white indicates no correlation.

Sleep PC1, derived from total Wake, NREM, and REM, captures the dominant axis of variation in overall amounts of sleep and wake across the 24-hour day. In fentanyl-treated mice, higher intake generally corresponded to negative correlations with Sleep PC1 (females: Week 3 active pokes, r = −0.55; males: Week 3 infusions, r = −0.54), reflecting a global reduction in sleep-wake amounts. In contrast, saline-treated mice showed positive correlations (females: Week 3 total infusions, r = 0.72; males: early abstinence active pokes, r = 0.81), consistent with greater behavioral engagement aligning with higher overall sleep-wake quantity. Together, these findings reveal a bidirectional, drug-dependent relationship between behavior and sleep, with fentanyl reversing the pattern observed in controls (Fig. 4B).

3.6. Linear models reveal sex- and fentanyl-related contributions to sleep principal components

We examined sleep- and IVSA-related principal components (PCs) across sex and group (Supplementary Table 3). Sleep PC1 explained 50.8% of the variance and Sleep PC2 explained 28.9% of the variance. IVSA PC1 explained 56.0% of the variance and IVSA PC2 explained 15.4% of the variance. Sleep PC1 showed group-specific associations with IVSA PC1: negatively correlated in saline animals (female: r = −0.45; male: r = −0.17) but positively correlated in fentanyl animals (female: r = +0.23; male: r = +0.29), indicating that fentanyl exposure alters the relationship between sleep and drug-taking behavior (Fig. 5A). In contrast, Sleep PC2 clustered by sex regardless of group, capturing sex differences (Fig. 5B).

Figure 5. PC1 captures treatment effects and PC2 captures sex differences in sleep-IVSA relationships.

Scatter plot showing relationships between (A) Sleep PC1 and IVSA PC1 and (B) Sleep PC2 and IVSA PC2 across sex (female, male) and treatment (saline, fentanyl). For PC1, group-specific regression lines indicate negative slopes in saline and positive slopes in fentanyl. For PC2, data cluster primarily by sex. (C) Bar plot showing standardized effect estimates (β) from linear models predicting Sleep PC1 and PC2 from IVSA PC1, IVSA PC2, sex, and treatment, illustrating how these factors contribute to variation in sleep-related principal components. Bars are colored by FDR significance (red = significant, grey = not significant), with panels split by factor type (sex, group). (D) Dot plot showing the absolute loadings of IVSA (left) and sleep variables (right) on PC1 (top) and PC2 (bottom). Each point represents a variable, colored and shaped by modality, and the dashed horizontal line at 0.3 marks a threshold for substantial loadings. This plot highlights which IVSA and sleep measures contribute most strongly to each principal component.

Linear modeling confirmed these patterns: Sleep PC2 was significantly influenced by sex (male vs. female: estimate = 2.16, p_adj = 0.0079), whereas Sleep PC1 showed a trend for group effects (saline vs. fentanyl: estimate = −2.30, p_adj = 0.068) (Fig. 5C). Together, these results indicate that Sleep PC1 reflects fentanyl-related changes in sleep-IVSA associations, while Sleep PC2 captures sex-dependent differences in sleep patterns.

3.7. PC loadings highlight key IVSA and sleep contributors to overall behavior

Analysis of absolute PC loadings identified the IVSA and sleep variables that contributed most strongly to overall behavior (Fig. 5D–E). For PC1, IVSA variables exceeding the 0.3 threshold included active responses, active inter-response interval, infusions, response rate, timeout responses, and inter-infusion interval, while sleep variables included dark-phase NREM, total NREM, total REM, dark-phase wake, and total wake (Fig. 5D). In contrast, PC2 was driven primarily by IVSA measures including inactive responses, latency to first active response, within-session escalation, and percent accuracy, together with light-phase sleep measures including NREM and wake (Fig. 5E). These loadings indicate that PC1 reflects overall IVSA engagement and general sleep/wake patterns. PC2, in contrast, reflects more specific aspects of operant performance, and with light-phase sleep changes.

4. Discussion:

Our study demonstrated that volitional fentanyl self-administration produced both acute and persistent, sex-specific disruptions in sleep architecture in mice. During drug-taking, fentanyl increased dark-phase NREM sleep, whereas abstinence was characterized by reduced REM sleep, and alterations in REM bout structure, indicating sustained alterations in sleep regulation across phases of the addiction cycle. Individual differences analyses revealed that REM sleep was most strongly associated with fentanyl-taking behavior, particularly in females. In saline-treated mice, operant engagement was positively associated with sleep measures, whereas fentanyl exposure reversed this relationship. Together, these findings indicate that fentanyl disrupts the normal coordination between drug-taking behavior and sleep-wake regulation.

Previous rodent studies using EEG/EMG recordings consistently show that opioids disrupt sleep during intoxication and withdrawal. In mice, fentanyl administered at the onset of the dark phase produced a biphasic effect, including increased wakefulness and reduced NREM during the first 6 hours, followed by a rebound in NREM during the next 6 hours. This rebound was attenuated by dopamine D2 receptor blockade in the nucleus accumbens core, highlighting a key role for accumbal dopaminergic signaling(Sharma, Parikh et al. 2024). In our study, sleep recordings were taken after IVSA sessions, thus initial wake-promoting effects may have subsided by that time, allowing the dark-phase NREM increases to be observed, consistent with Sharma et al. (2024). However, other studies also report increased wakefulness or reduced NREM sleep after opioid exposure. For example, acute oral morphine self-administration increased wakefulness, while chronic morphine selectively elevated dark-phase wake(Eacret, Manduchi et al. 2023). Similarly, injections of fentanyl, morphine, or buprenorphine reduced both NREM and REM sleep and altered EEG power in male mice(O’Brien, Locklear et al. 2021). In contrast, during withdrawal, opioids continue to disrupt sleep but in different ways. Tisdale and colleagues reported that withdrawal from escalating morphine injections increased dark-phase NREM and REM, with females showing greater sleep fragmentation and reduced NREM delta power(Tisdale, Sun et al. 2024). Using piezo sleep chambers and twice-daily injections of fentanyl, Gamble and colleagues reported that males showed reduced REM during withdrawal(Gamble, Miracle et al. 2024), consistent with our observation of reduced REM during abstinence across sexes. Together, these findings suggest that opioid-induced sleep disruption depends on drug type, route of administration, and sex.

These fentanyl-induced sleep changes raise questions about the neural circuits and molecular mechanisms underlying opioid effects on sleep and circadian regulation. In mice, chronic morphine via oral self-administration activates μ-opioid receptor-expressing neurons in the paraventricular nucleus of the thalamus (PVT) and alters circadian gene expression, while inhibition of these neurons reduces morphine-induced wakefulness without affecting baseline sleep(Eacret, Manduchi et al. 2023). As a potent μ-opioid receptor agonist, fentanyl may engage overlapping PVT circuits, which could contribute to its sleep effects; however, because we observed increased NREM sleep, the specific impact on NREM versus wakefulness remains unclear. Additionally, our finding of reduced and fragmented REM sleep during fentanyl abstinence may contribute to hypothalamic-pituitary-adrenal (HPA) axis dysregulation. Consistent with this, sleep restriction during opioid abstinence altered diurnal and stress-induced adrenocorticotropic hormone and corticosterone responses in a sex-specific manner, with males showing blunted stress responses and females failing to return to baseline post-stress(Raff, Glaeser et al. 2023). Chronic sleep restriction alone, independent of opioid exposure, also disrupts basal hormone levels, adrenal responsiveness, and feedback control in a sexually dimorphic pattern(Everson, Szabo et al. 2025). These findings are particularly relevant to our model, where reduced REM during abstinence could promote HPA axis dysfunction, a factor linked to greater stress vulnerability and increased relapse risk in humans(Brown, Wisniewski et al. 2006, Daughters, Richards et al. 2009). Collectively, these observations suggest that fentanyl-induced changes in accumbal dopamine, arousal circuits, and stress signaling may contribute to the sex- and drug phase-specific sleep alterations we observed. These preclinical findings may help explain the persistent sleep disruptions observed in individuals with OUD.

Emerging clinical evidence demonstrated that individuals with OUD experienced profound and persistent disruptions in sleep and circadian rhythms(Logan, Hasler et al. 2018, Eacret, Veasey et al. 2020, Eckert and Yaggi 2022, Langstengel and Yaggi 2022, Mehranfard, Ghasemi et al. 2025). OUD is characterized by severe and persistent sleep disturbances, with many patients meeting criteria for clinical sleep disorders during active opioid use, withdrawal, and while receiving medications for OUD(Hartwell, Pfeifer et al. 2014, Baldassarri, Beitel et al. 2020, Greenwald, Moses et al. 2021, Huhn and Finan 2022, White, Eglovitch et al. 2024). Notably, a recent study by Zhang and colleagues found that individuals with OUD, including those receiving opioid agonist treatment, exhibited greater sleep-wake irregularity, which was associated with longer drug use histories, reduced daytime light exposure, and increased dominance of default mode network (DMN) brain states (DMN describes set of brain regions that are most active when an individual is at rest) (Zhang, Manza et al., 2025). Chronic opioid exposure further exacerbated these disturbances(Shi, Zhao et al. 2007), and sleep and circadian disruption during treatment and abstinence were associated with heightened craving and increased risk of relapse(O’Connor and Fiellin 2013, Lydon-Staley, Cleveland et al. 2017, Fathi, Yoonessi et al. 2020, Sun, Wang et al. 2023).

Among individuals undergoing inpatient opioid detoxification, greater insomnia severity over the prior two weeks, but not prior-night sleep duration, predicted larger stress-induced increases in craving and negative affect(Bichon, Bailey et al. 2025). These findings suggest that chronic sleep disruption, rather than acute nightly sleep loss, is particularly relevant to vulnerability in OUD. Consistent with this, we observed persistent reductions in REM sleep during abstinence following fentanyl self-administration in mice, indicating that opioid exposure produces enduring alterations in sleep architecture that persist beyond drug use and this may contribute to relapse vulnerability. Similarly, a retrospective study of 154 adults with OUD found progressive declines in sleep across the lifespan, with the most persistent disturbances in those with co-occurring insomnia, chronic pain, female sex, or multiple treatment episodes(Ellis, Mayo et al. 2022). These findings highlight that sleep problems in OUD are common, often persistent, sex-biased, underdiagnosed, and may both precede and exacerbate opioid use, making them a targetable factor for treatment and prevention. Supporting this, sleep data from 1,905 individuals with OUD receiving treatment across 70 US programs showed that greater insomnia severity during and after treatment predicted a higher risk of return to opioid use and nonfatal overdose. Notably, insomnia severity at treatment intake predicted relapse within one month, and persistent insomnia remained a strong predictor of opioid use at 1, 3, and 6 months post treatment(Hochheimer, Ellis et al. 2025). Together, these findings highlight insomnia as a critical and potentially modifiable risk factor for poor outcomes in OUD.

While our study had many advantages, including using IVSA (the gold standard model), examining the understudied effects of fentanyl on sleep, and including both sexes, a limitation is that sleep was measured with the noninvasive PiezoSleep system rather than EEG/EMG. However, this approach allowed continuous data collection across every day of drug taking and abstinence, which would be challenging to access otherwise. Importantly, one limitation of measuring sleep by either procedure is the necessity for singly housing the animals, which may cause stress. However, more recent studies have shown nominal, and some positive (males only), effects of singly housing C57BL/6 mice (Smolensky, Zajac-Bakri et al. 2024, Davies, Jackson et al. 2025). While saline self-administration is uncommon, prior work demonstrates that mice can exhibit operant responding for saline and associated cues under specific conditions (David, Polis et al. 2001, Schramm-Sapyta, Olsen et al. 2006, Olsen and Winder 2012, López, Johnson et al. 2021). Specifically, using an FR1 schedule of reinforcement with nose ports may increase saline-directed responding, potentially driven by cue-associated reinforcement. Additionally, procedural factors such as testing during the active (dark) phase, as in the current study, may elevate overall activity and contribute to increased saline responding compared to studies conducted during the inactive phase (Nelson, Bumgarner et al. 2021). Thus, saline responding in our study may reflect sensitivity to operant contingencies and associated cues rather than drug-seeking behavior. Future studies incorporating molecular approaches will be important to elucidate the mechanisms underlying this behavior.

In summary, our findings demonstrate that volitional fentanyl self-administration disrupts sleep in a sex- and phase-specific manner, increases NREM sleep during drug taking, produces persistent alterations in REM sleep during abstinence, and reverses the correlation between sleep and operant behavior compared to saline controls. These results highlight the bidirectional relationship between opioid use and sleep disruption and support the growing view that sleep and circadian dysfunction are not merely consequences of opioid exposure, but integral components of OUD pathology. Targeting sleep and circadian systems, through behavioral, pharmacological, or chronotherapeutic interventions, may therefore represent a promising and underutilized strategy to reduce relapse risk and improve treatment outcomes in opioid use disorder.

Highlights.

1. Fentanyl IVSA increases dark-phase NREM sleep in mice

2. Abstinence causes persistent REM reduction and fragmentation

3. REM sleep correlates with fentanyl intake, especially in females

4. Fentanyl reverses the normal, positive correlation between sleep and IVSA behavior observed with saline.

5. Sleep disruption persists into abstinence and may drive relapse vulnerability

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.