Telemetry surgeries.

Anesthetized rats were implanted with the wireless contained telemetry device, with 4 biopotential leads: 2 EMG leads that comprise one EMG channel and 2 EEG leads that comprise one channel. For each implant, the EMG leads were threaded through the cervical trapezius muscle via a small incision made with a 20-gauge needle and was anchored to the muscle using non-dissolvable silk sutures. The EEG leads were individually secured to two skull screws (Plastics One, cat #00–80 x 1/16; 1.6mm long) that contacted dura. The negative EEG lead was secured to the screw at +2.0mm anteroposterior and at lateral left 1.5mm, and the positive EEG lead was secured to the screw at -7.0mm anteroposterior and lateral right -1.5mm–all relative to Bregma. The screw plus lead assemblies were held in place on the skull with dental cement (Ortho-Jet Package; Lang Dental).

iPrecio drug pump surgeries.

Immediately after implanting the telemetry device, iPrecio drug pumps were subcutaneously implanted. The pump tubing, from which drug is released, was cut to 2.5–3.0cm prior to implantation. A subcutaneous pocket was created at 3-4cm posterior to the scapula and 2cm lateral and parallel to the spine, and the pre-programmed iPrecio pump was inserted. To secure the pump within the pocket, a suture was threaded through the underlying external oblique muscle and then through a suture anchor located on the pump. After surgery, rats were treated one time with the non-steroidal anti-inflammatory drug (NSAID) Ketoprofen (5 mg/kg, IP) and individually housed in the sleep room.

Oxycodone administration.

Prior to implantation, iPrecio pumps were programmed using the iPrecio Management System and Software (IMS-200; ALZET) to infuse either saline or oxycodone twice a day for two hours each (ZT0-2 and ZT12-14). For the first several days after surgery, the pumps infused 0.9% saline for 2h. Fig 1A shows a schematic of the experimental design. Telemetry data from the last 3 days of saline infusions were used as baseline. Following baseline, saline was removed from the pumps and replaced with oxycodone (0.5 mg/kg). The timeline for escalating oxycodone doses included a total of 28 oxycodone infusions (ZT0-2 and ZT12-14 each day). On oxycodone day 1 (Oxy d1), rats received saline in the morning (ZT0-2) and oxycodone (0.5 mg/kg) in the evening (ZT12-14). On the last drug day, Oxy d15, rats received oxycodone (8.0 mg/kg) in the morning (ZT0-2) and no infusion in the evening (ZT12-14). Between Oxy d1 and Oxy d15, rats received 24 additional infusions of oxy (0.5 mg/kg), 6 infusions of oxy (1.0 mg/kg), 6 infusions of oxy (2.0 mg/kg), 6 infusions of oxy (4.0 mg/kg), and 5 infusions of oxy (8.0 mg/kg), with the 6^th oxy (8.0 mg/kg) infusion occurring on Oxy d15 (ZT0-2). For all cohorts except cohort 2, infusions were delivered at rates of 15μl/h (30 μl infused) for the first 4 infusions and 30μl/h (60 μl infused) for all remaining infusions. Cohort 2 received saline during baseline and the oxy 0.5mg/kg infusions at 7.4μl/hr, oxy 1.0mg/kg at 14.8μl/hr and all remaining infusions at 29.8μl/hr. Separate rats received saline throughout the 14-d pump delivery regimen and served as controls for any effects of surgeries, interoceptive effects of infusions, isoflurane exposures, and aging/weight gain on sleep. In addition, saline rats controlled for telemetry implant and receiver stability over the course of the experiment.

The iPrecio pumps were emptied and refilled with the appropriate concentration of oxycodone (or saline) 4 times, always between ZT7-12. Oxycodone concentrations were calculated based on individual rat weights, the volume infused, and the rate of infusion. To refill the pumps, rats were briefly anesthetized with 4–5% isoflurane, and a 26-gauge syringe was used to withdraw all fluid in the reservoir and add new drug via a port on the pump that could be felt through the skin. The volume of drug withdrawn during a pump change served as a control that the pumps were delivering the expected volume of drug.

Oxycodone withdrawal.

To demonstrate that the escalating-dose oxycodone regimen produced dependence, rats were injected with the opioid receptor antagonist, naloxone (1.0 mg/kg, SC), or saline, within an hour (9–10 AM) (after the last oxycodone (8.0 mg/kg) infusion ended on Oxy d15 [referred hereafter as withdrawal day 1 (W1)] to assess naloxone-precipitated somatic withdrawal signs (with the exception of rat A36 who received a saline injection at 11 AM). Immediately after naloxone (or saline) injections, rats were videotaped in their home cages for at least 35 minutes, and minutes 20–35 of the videos were subsequently scored for withdrawal signs by a trained experimenter, as previously described [27]. Briefly, videos were scored for somatic withdrawal signs including ptosis, flattened posture, rats putting their jaw/snout to the surface or bedding of the cage, and stereotyped head bobs (where the rat is sitting down and sniffs the air or moves only their head). Scoring was standardized such that every 15 seconds, the scorer would glance at the video and note which behavior a rat was doing at that moment. To derive a Total Withdrawal score, the instances of each recorded behavior were summed over the 15 minutes. Because rats received either naloxone or saline injection on W1, we did not consider sleep data collected on that day to cleanly represent spontaneous withdrawal. Rather, ZT0 on W2 marked the start of spontaneous oxycodone withdrawal in our study. 11 out of the 12 oxycodone rats had their W1 behavior analyzed (see Fig1_Data in S1 File).

Sleep scoring.

Sleep was manually scored by a trained experimenter in 10-second epochs by visually inspecting the EEG and EMG signal combined with a spectral analysis of the EEG waveform using NeuroScore software. Using a standard, objective set of rules based on a combination of behavior and EEG events, sleep states were categorized into wake, rapid eye movement (REM) sleep, and non-REM (NREM) sleep (Missig et al., 2018). NREM sleep was characterized by the presence of high amplitude and low frequency waves in the EEG combined with relatively little activity in the EMG. Wake was characterized by low amplitude EEG combined with relatively high activity in the EMG and locomotor activity. REM sleep was characterized by the dominance of theta waves (4-8Hz). Sleep stage scoring was done by the same experimenter who handled the 14-d drug infusion phase of the experiment.

Spontaneous oxycodone withdrawal transiently flattens sleep-wake rhythms.

To characterize the effects of spontaneous oxycodone withdrawal on the diurnal rhythmicity of sleep architecture observed during baseline recording days, we plotted the percent of total time spent in NREM, REM, and Wake (Fig 2A–2C, respectively) in 3-h bins covering the last 3 days of baseline recordings (saline infusions, B4, B5, B6) and spontaneous oxycodone withdrawal days 2–8 (W2-W8). Withdrawal day 1 (W1) is not included in our spontaneous withdrawal results because rats received their last 2-h infusion of oxycodone and an injection of either naloxone or saline on the morning of W1 (see Methods). For all sleep states, we observed typical diurnal rhythmicity during B4-B6, followed by a marked flattening (loss of rhythmicity) during early withdrawal (~W2-W4; Fig 2A–2C) and a return towards baseline patterns by W8. To quantify the effects of spontaneous oxycodone withdrawal on diurnal rhythmicity during NREM, REM, and Wake, we compared the magnitude and timing of several sleep stage features during W2-8 to the average of the last 3 baseline days (BL_avg). First, we computed the diurnal index (DI), defined here as the peak-to-trough amplitude in the 3-h bin values on a particular day, normalized to each individual rat’s average peak-to-trough amplitude over the last 3 baseline days (BL_avg). A decrease in DI relative to BL_avg indicates a reduction in the magnitude of diurnal rhythmicity (i.e., flattening). For the % of time in each sleep state, Friedman’s Tests show that the DI is significantly reduced compared to BL_avg: [NREM, Q = 44.95, p<0.0001, Fig 2D, left; REM, Q = 31.44, p<0.0001, Fig 2E, left; Wake, Q = 45.18, p<0.0001, Fig 2F, left). To control for the possibility that rats injected with naloxone on W1 (1.0 mg/kg, SC; N = 4) would show different spontaneous oxycodone withdrawal sleep effects compared to rats injected with saline on W1 (1.0 ml/kg, SC; N = 7), we conducted 2-way ANOVAs with repeated measures on withdrawal day for W2-W8 on naloxone vs saline rats. There were no Treatment x Day interactions or Treatment main effects for any sleep stage (S1A), indicating no effect of W1 naloxone injection on subsequent spontaneous oxycodone withdrawal effects on sleep architecture.

To determine whether the effects of oxycodone withdrawal on sleep/wake occur primarily during lights on or off, we quantified the average percent time spent in NREM, REM and Wake for ZT0-12 and for ZT12-24 (Fig 2D–2F, middle panels) and compared W2-W8 to BL_avg. Overall, we find that during early withdrawal the % time asleep decreases, whereas the % time awake increases, during lights on (NREM, Q = 22.24, p = 0.0023, Fig 2D, middle; REM, Q = 40.55, p<0.0001, Fig 2E, middle; Wake, Q = 28.83, p = 0.0002, Fig 2F, middle). In contrast, there is a general increase in % time asleep and a decrease in % time awake during lights off (NREM, Q = 33.25, p<0.0001, Fig 2D, middle gray; REM, Q = 30.18, p<0.0001, Fig 2E, middle gray; Wake, Q = 33.36, p<0.0001, Fig 2F, middle gray). Together with the DI data, these findings indicate that the most robust effects of spontaneous withdrawal from 14 days of chronic, escalating-dose oxycodone occur during W2 and W3. Further, there is an association between loss of diurnal rhythmicity and insomnia-like reductions in time asleep and subsequent increases in lights off (active period) time asleep.

Since our data suggest a robust flattening of diurnal rhythmicity for each sleep stage on W2 with day-by-day incremental improvements toward baseline, we performed a more granular analysis on the % time in each sleep stage on W2 and W8 by examining the data in 3-h bins and comparing the results to BL_avg (Fig 2D–2F; right panels). As expected, the % time in each sleep stage on W2 is relatively flat from ZT0-ZT24, with little difference between lights on and off. On W8, however, recovery towards the rhythmicity of BL_avg is observed during lights on, but not lights off. Indeed, similar to W2, the % time in REM and Wake on W8 during several 3-h bins in lights off is significantly different from BL_avg. These results depend on significant interactions (Treatment day × ZT hour) from 2-way repeated measures ANOVA tests followed by Dunnett’s multiple comparison tests: REM, F_{(4.995, 49.95)} = 13.21, p<0.0001, Fig 2E, right panel; Wake, F_{(3.339, 33.39)} = 11.50, p<0.0001, Fig 2F, right panel.

Like effects on sleep architecture rhythmicity, spontaneous oxycodone withdrawal reduces the magnitude of activity and temperature diurnal rhythmicity (Fig 3A and 3B, respectively), with the greatest effects occurring on W2 and W3 and subsequent recovery towards BL_avg by W8. For temperature and activity, Friedman’s Tests show that the DI is significantly reduced compared to BL_avg: Temp, Q = 34.85, p<0.0001, Fig 3C, left panel; Activity, Q = 48.52, p<0.0001, Fig 3D, left panel. In contrast to the effects of oxycodone withdrawal on sleep; effects on activity and temperature rhythmicity occurred primarily during lights off. This is particularly evident in Fig 3C and 3D, right panels, in which 2-way repeated measures ANOVAs show significant Treatment day × ZT hour interactions (Temp, F_{(3.293, 32.93)} = 7.381, p<0.001, Fig 3C, right panel; Activity, F_{(2.742, 27.42)} = 7.953, p<0.001, Fig 3D, right panel), and Dunnett’s post hoc tests show that significant pairwise comparisons occur during lights off.

Both sleep and wake show decreased bout durations during early spontaneous oxycodone withdrawal.

Above, we demonstrate that during early spontaneous withdrawal there are overall reductions in the amplitudes of diurnal rhythms for each sleep stage, temperature, and activity, which result in overall changes in the amount of sleep/wake during lights on and off. We next investigated whether oxycodone withdrawal affects sleep stage switching, either through increased or decreased average sleep/wake bout durations. Hypnograms (experimenter-scored sleep stages plotted over time; Fig 4A) are shown for B4-B6 and W2-W8 from a representative oxycodone-treated rat. The hypnograms show marked differences between baseline days and W2 and 3. Qualitatively, the long-duration wake bouts predominant during lights-off decrease in length on W2, replaced with an increased number of sleep bouts, suggesting disrupted sleep/wake architecture on W2 and W3 relative to baseline.

To characterize the degree to which spontaneous oxycodone withdrawal alters sleep/wake architecture, we quantified bout number and bout length (in seconds) for each sleep stage during either lights-on (Fig 4B–4D; left panels) or lights-off (Fig 4B–4D; right panels) and compared these to their respective BL_avgs. Consistent with what is clear to the eye from Fig 4A, the most robust effects are seen on W2 and W3 which include increased Wake bouts during lights on and off and decreased Wake bout durations during lights off (Fig 4D). This is demonstrated with significant Friedman’s tests for counts and durations (Wake bout counts, lights on: Q = 49.17, p<0.0001, Fig 4D, left; lights off: Q = 49.52; p<0.0001, Fig 4D, right; Wake bout duration, lights off: Q = 48.94, p<0.0001, Fig 4D, left) During lights on, the number of Wake bouts increases compared to BL_avg, but the bout length does not change. Since the number of NREM and REM bouts do not decrease during lights-on, the increase in Wake bouts is enabled by the decrease in NREM and REM bout length (B, C, left panels). Friedman’s tests show that NREM and REM bout lengths decrease: for NREM and REM during lights on (NREM, Q = 58.45; p<0.0001 Fig 4B, left; REM, Q = 30.82, p<0.0001, Fig 4C, left) Interestingly, increased Wake bout number ends on W4, yet NREM bout durations remain significantly decreased until W6. During lights off, the mean number of Wake bouts dramatically increases from 56.61 (±2.89 SEM) at BL_avg, Fig 4C to 107.91 (±8.17 SEM) on W2 and 85.64 (±4.99 SEM) on W3, with concurrent decreases in bout length from 556.80s (±31.61 SEM) at BL_avg to 189.59s (±14.75SEM) on W2 and 277.99s (±26.62SEM) on W3. The increased bouts and decreased bout durations for Wake staging during lights off are accompanied by significant increases in the number of NREM and REM bouts, as shown by significant Friedman’s tests (NREM, lights off: Q = 47.03; p<0.0001,; p<0.005, Fig 4B, right REM, lights off: Q = 24.20; p = 0.0010, Fig 4C, right). To summarize, oxycodone withdrawal results in profound arousal/sleep state switching and decreased bout durations. These changes suggest both severe disruption of normal rat sleep architecture during early oxycodone withdrawal and differential effects on the neural circuitry that regulates sleep/wake initiation maintenance.

To control for the possibility that oxycodone withdrawal-induced changes to sleep/wake bout architecture are modulated by the acute naloxone injections on W1, we conducted 2-way ANOVAs with repeated measures on withdrawal day (W2-W8) for NREM bout # and bout duration in naloxone vs saline rats. There were no Treatment x Day interactions or main effects of Treatment for any measurement (S1B), indicating no effect of W1 naloxone injection on NREM sleep architecture.

Spectral power and aperiodic EEG aperiodic slope are differentially and temporally altered during spontaneous oxycodone withdrawal.

Beyond sleep architecture and diurnal rhythms, it is vital to understand if, and how, spontaneous oxycodone withdrawal affects the magnitude, directionality, and temporal dynamics of functional network activity underlying sleep. Accordingly, we examined EEG spectral dynamics for NREM, REM, and Wake across baseline B4-B6 and withdrawal W2-W8 days. We first derived estimates of relative spectral power for the canonical frequency bands delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–50 Hz). Spontaneous oxycodone withdrawal had the most robust effects on spectral power and aperiodic 1/f slopes during NREM. As such, our results and discussion focus on NREM, although data from REM and Wake is presented in the supplementary data in S1 File.

Spontaneous oxycodone withdrawal transiently flattens the diurnal rhythmicity of NREM relative spectral power, but not that of REM or Wake. Given the flattening of baseline diurnal rhythms within sleep architecture, we next analyzed EEG relative spectral power to assess any effects observed concurrently during spontaneous oxycodone withdrawal. Qualitatively, NREM relative power measured over baseline (B4-B6) shows a clear rhythmicity [52] for each canonical frequency band (Fig 5), with the exception of theta (Fig 5B). For delta, relative power is highest at the beginning of lights-on (ZT0), decreases until ZT12 where it is at its lowest, and increases again within ZT12-24. In contrast, alpha, beta, and gamma show the opposite pattern, with the lowest relative power occurring at ZT0. During oxycodone withdrawal, NREM rhythms of power appear to flatten before gradually recovering to baseline periodic structure by W8. To determine if the rhythmicity of power was affected by oxycodone withdrawal, we quantified the DI for NREM across all frequency bands (Fig 5A–5E; left panels). A Friedman’s test revealed significant decreases to the DI for every frequency band; including theta, which had a qualitatively unclear rhythm (Delta, Q = 34.54, p<0.0001; Fig 5A; Theta, Q = 18.21, p<0.0111; Fig 5B; Alpha, Q = 38.67, p<0.0001; Fig 5C; Beta, Q = 37.71, p<0.0001; Fig 5D; Gamma, Q = 28.88, p = 0.0002; Fig 5E). Thus, spontaneous oxycodone withdrawal flattens NREM rhythms for all frequency bands.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Spontaneous oxycodone withdrawal transiently eliminates diurnal changes in NREM depth.

The mean relative power for NREM is plotted in 3-h bins (±SEM) for (A) Delta (1-4Hz), (B) Theta (4-8Hz), (C) Alpha (8-12Hz), (D) Beta (12-30Hz) and (E) Gamma (30-50Hz) frequency bands throughout baseline days 4, 5, and 6 (B4, B5, B6; blue line, circles) and spontaneous withdrawal days 2–8 (W2-W8; red line, diamonds). Vertical green bars between B6 and W2 represent the 14-d escalating oxycodone dose regimen and W1. Grey shaded regions represent ZT12-24 (lights off), and unshaded regions represent ZT0-12 (lights on). Oxycodone withdrawal has a global effect on relative power, rather than modulating specific frequency ranges. N = 8 rats. Underlying data is in Fig5_Data in S1 File.

https://doi.org/10.1371/journal.pone.0312794.g005

Qualitatively, REM and Wake relative power measured over baseline B4-B6 days appear to show subtle rhythmicity at individual frequency bands (S3, S5, Figs in S1 File, respectively), but more detailed analyses of these potential rhythms were not performed, as they are beyond the scope of this initial study. Furthermore, DI analyses did not reveal any significant differences between relative power at BL_avg compared to W2-W8 across frequency bands for either sleep stage (S4 Fig in S1 File, REM; S6 Fig in S1 File, Wake, left panels).

Early spontaneous oxycodone withdrawal has global effects on NREM relative power across frequencies, with minimal effects on REM and Wake. We next quantified the mean NREM, REM, and Wake relative power for BL_avg and W2-W8 during either lights-on or lights-off (Fig 6, NREM; S4 Fig in S1 File, middle panels, REM; S6 Fig in S1 File, middle panels, Wake) and performed Friedman’s Tests followed by Dunn’s post hoc tests for significance. For NREM, spontaneous oxycodone withdrawal shows a continuous gradient in power changes with increasing frequency during lights-on, but sparse changes in lights-off. Compared to BL_avg, relative power during NREM delta is significantly decreased for lights on (Q = 21.04, p = 0.00371; Fig 6A, middle white). Between 4–12 Hz (theta and alpha) there is a switch in directionality such that relative power in higher frequencies (12–50 Hz, beta and gamma) is significantly increased for lights on (beta, Q = 27.13, p = 0.0003; Fig 6D, middle white; gamma, Q = 39.67, p<0.0001; Fig 6E, middle white). During lights-off, NREM relative power shows a few significant changes compared to BL_avg. Interestingly, these (moderate) changes peak on later withdrawal days than those observed for % time in sleep states, sleep/wake bout architecture disruption, and diurnal rhythmicity. Specifically, the peak increase in NREM relative power compared to BL_avg during lights-off occurs on W5 for delta (1–4 Hz; Q = 14.79, p = 0.0388; p = 0.0154 at W5, Dunn’s test, Fig 6A, middle gray panel). The strong effect of oxycodone withdrawal on NREM relative power magnitude during lights-on compared to lights-off can best be seen in the right panels of Fig 6. In particular, spontaneous oxycodone withdrawal dramatically decreases delta power and increases beta and gamma power during lights-on, with the maximum effects on W2 occurring between ZT0-9. This is demonstrated with 2-way repeated measures ANOVAs and significant 3-h bin × Treatment Day interactions (delta, F_{(2.085, 14.60)} = 9.130, p = 0.0025, Fig 6A, right panel; beta, F_{(1.838, 12.87)} = 11.92, p = 0.0014, Fig 6D, right panel; gamma, F_{(3.081, 21.57)} = 13.66, p<0.0001, Fig 6E, right panel). This suggests that oxycodone withdrawal has long-lasting effects on oscillatory activity when transitioning into the sleep period (lights-on). Across W2-W8, oxycodone withdrawal did not have robust or consistent effects on REM or Wake power during lights-on/off, as analyzed and shown in S4 Fig in S1 File, REM and S6 Fig in S1 File, Wake. A finding shared by all three sleep stages is that relative power is at BL_avg levels by W8 across all frequency bands, broadly similar to % time in sleep stages and sleep/wake bout architecture (Figs 2 and 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Spontaneous oxycodone withdrawal initially flattens NREM power rhythms, decreases low frequency and increases high frequency power.

For relative power during NREM, the mean (±SEM) Diurnal Index (DI; left panels), mean (±SEM) relative power per frequency band for ZT0-12 or ZT12-24; center panels, and the mean (±SEM) relative power per frequency band in 3-h bins for BL_avg, W2, and W8 (right panels) are shown for (A) Delta 1-4Hz, (B) Theta (4-8Hz), (C) Alpha (8-12Hz) (D) Beta (12-30Hz), (E) Gamma (30-50Hz) frequency bands. Non-parametric Friedman with Dunn’s post hoc tests (left and center panels), and two-way repeated measures ANOVA with Dunnett’s post hoc tests (right panels) compared W2-W8 to B_avg. ***p<0.001.**p<0.01, *p<0.05. N = 8 rats. Underlying data is in Fig6_Data in S1 File. Abbreviations: ZT, zeitgeber time; BL_avg, Baseline average; W, withdrawal.

https://doi.org/10.1371/journal.pone.0312794.g006

To control for the possibility that oxycodone withdrawal-induced changes to NREM power are modulated by the acute naloxone injections on W1, we conducted 2-way ANOVAs with repeated measures on treatment day (BL_avg, W2, and W8) for NREM relative power at Delta and Gamma frequency ranges in 3-h bins in naloxone vs saline rats. For Gamma, we found significant Treatment x Day interactions: Gamma: F_(21,84) = 2.473, p = 0.0019 (S1C Fig in S1 File). However, Bonferroni’s posthoc tests showed no significant pairwise effects between naloxone and saline treated rats, indicating no effect of W1 naloxone injection on subsequent spontaneous oxycodone withdrawal effects on this endpoint.

Spontaneous oxycodone withdrawal transiently flattens NREM aperiodic slopes at low frequencies, with minimal effects on REM and Wake. The gradient of spectral power changes observed in lights-on NREM during spontaneous withdrawal raises the possibility that the entire spectrum is uniformly affected rather than a series of variable band-specific changes. Specifically, these changes are consistent with a generalized “tilt” in the aperiodic spectrum around theta/alpha ranges. We therefore explicitly assessed this possibility by quantifying the aperiodic piece-wise low frequency (LF, 1-4Hz) and high frequency (HF, 15-50Hz) spectral slope and their changes during the experiment. For visualization, we show example NREM relative spectra (in log-log) taken from a rat at baseline (BL_avg), W2, and W8 during lights-on (ZT0-12) and during lights-off (ZT12-24) (Fig 7A, top panels) with a close-up of the HF and LF regions (Fig 7A, lower panels). For the LF lights-on (lower, left), we observe a distinct flattening of the spectral slope at W2 (red), pivoting down from a point at ~4Hz. For the HF-lights on and lights off (lower, right), we also see a flattening in the W2 (red) slope, but this time pivoting up from the lower end at ~15 Hz. To support the validity of separately analyzing LF and HF portions of the PSD to determine spectral slopes, we calculated 1/f slopes for NREM PSDs at 3 LF frequency ranges (1–4 Hz, 1–10 Hz, and 1–15 Hz; S7A-S7C Fig in S1 File) and at 3 HF frequency ranges (15–50 Hz, 15–30 Hz, and 15–45 Hz; S8A-S8C Fig in S1 File). For both the LF and HF ranges, the DI, 1/f slopes, and 3-h bin data for BL, W2, and W8 were respectively similar, suggesting a general robustness of our findings to the aperiodic frequency range.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. NREM 1/f slopes calculated from Power Spectrum Densities (PSDs) have a robust nocturnal rhythm at low frequencies that are initially flattened during spontaneous oxycodone withdrawal.

(A, top panel) The average NREM PSD normalized to each rat’s total power from 0.5-50Hz plotted on a log-log scale for lights-on (ZT0-12 unshaded) and lights-off (ZT12-24; shaded) for BL_avg (blue), W2 (red), and W8 (pink). (A, bottom panels) Zoomed-in low frequency (LF, 1–4 Hz) and high frequency (HF, 15–50 Hz) portions of the NREM PSD during lights-on and lights-off are plotted and used to calculate LF and HF 1/f slopes. The mean NREM 1/f slope (±SEM) per 3-h bin is plotted for (B) low frequency and (C) high frequency ranges across baseline days 4, 5, and 6 (B4, B5, B6; blue line, circles) and spontaneous withdrawal days 2–8 (W2-W8; red line, diamonds). Vertical green bars between B6 and W2 represent the 14-d escalating oxycodone regimen and W1. For both LF (D) and HF (E) NREM 1/f slopes, the mean (±SEM; including individual rat values) Diurnal Index (left panels), 1/f slopes (ZT0-12 or ZT12-24; center panels), and 3-h bin data for BL_avg, W2, and W8 (right panels) are shown. Non-parametric Friedman tests with Dunn’s post hoc tests (left and center panels), and two-way repeated measures ANOVA with Dunnett’s post hoc tests (right panels) compared W2-W8 to B_avg. ***p<0.001.**p<0.01, *p<0.05. N = 8 rats. Underlying data is in Fig7_Data in S1 File. Abbreviations: ZT, zeitgeber time; BL_avg, Baseline average; W, withdrawal.

https://doi.org/10.1371/journal.pone.0312794.g007

To characterize aperiodic spectral dynamics for baseline and withdrawal days, we computed LF and HF slopes in 3-h bins for each sleep stage (Fig 7B and 7C, NREM; S9A, S9B Fig in S1 File, REM; S10A, S10B Fig in S1 File, Wake). As with relative power, we found that the most robust effects of oxycodone withdrawal on aperiodic activity occurred during NREM. Interestingly, the LF aperiodic signal (Fig 7B, blue) has a robust, diurnal rhythm at baseline that is similar in shape and directionality to the NREM periodic alpha, beta and gamma rhythms (Fig 5C–5E). To quantify rhythmicity changes to aperiodic slope at either LF or HF, we calculated the DI across all sleep stages by comparing W2-8 to BL_avg (Fig 7D and 7E, NREM; S9C, S9D Fig in S1 File, REM; S10C, S10D Fig in S1 File, Wake; left panels). For NREM, oxycodone withdrawal significantly decreases the LF slope DIs (Friedman’s test: Q = 22.46, p = 0.0021, Fig 7D, left). HF slope DIs did not change during withdrawal for NREM (Fig 7E, left). REM and Wake slopes appear highly variable and complex (S9, S10 Figs in S1 File), and quantification of slope DIs also revealed no significant differences compared to BL_avg for Wake (S10C, S10D Fig in S1 File, Wake, Figs, left panels). although there are some differences in DIs for HF REM slopes (Friedman’s test, Q = 27.92, p = 0.0002, S9D Fig in S1 File, REM, left panel).

To quantify the daily changes in slopes during withdrawal, we plotted NREM, REM, and Wake LF and HF slopes for BL_avg and W2-W8 separately for lights-on and lights-off (Fig 7D and 7E, NREM; S9C, S9D Fig in S1 File, REM; S10C, S10D Fig in S1 File, Wake; middle panels). We found significantly different slope magnitudes during withdrawal for all sleep stages, with the greatest effects in NREM. For NREM, the LF slope is significantly closer to 0 (indicates the slope is becoming flatter) on W2-W5 during lights-on when compared to BL_avg (Q = 30.71, p<0.0001; Fig 7D, middle white) but not during lights-off. This suggests that oxycodone withdrawal prevents normal light-induced changes to LF aperiodic activity. The effects of withdrawal on lights-on sleep metrics are the most prolonged for NREM. During withdrawal, NREM HF 1/f slopes are also significantly flattened (approaching 0) during lights on (Q = 29.88 p = 0.0001; Fig 7E, middle white), but this effect is only on W2. For REM, slopes are significantly different compared to BL_avg during lights-on and lights-off for LF (lights-on: Q = 14.71, p = 0.0399; S9C Fig in S1 File, middle white; lights-off: Q = 24.50, p = 0.009; S9C Fig in S1 File, middle gray), and during lights on for HF (lights on: Q = 23.92; p = 0.0002, S9D Fig in S1 File, middle white). Slopes at LF become steeper, unlike those for NREM. We report these findings, however given the low percent time in REM, additional data collection or experiments targeting REM-specific pathways may be required confirm these results.

Aperiodic activity follows the same patterns as those of power, sleep/wake %, activity, and temperature wherein oxycodone withdrawal effects are strongest at W2 before gradually returning to BL_avg levels by W8. Despite absolute (ZT0-24), or lights-on (ZT0-12) and lights-off (ZT12-24) levels returning to normal, we have repeatedly observed time-of-day effects on W8. This suggests that certain structural changes to behavior are longer lasting than others, and recovery from oxycodone withdrawal may not be complete by W8. As such, we determined the magnitude and time-of-day dependence for slopes at each sleep stage on BL_avg, W2, and W8 in 3-h bins (Fig 7D and 7E, NREM; S9C, S9D Fig in S1 File, REM; S10C, S10D Fig in S1 File, Wake; right panels). For NREM, HF slopes are significantly flattened during both lights-on and lights-off on W2 (Fig 7E, middle panels). When W2 slopes are calculated for 3-h bins, a more detailed picture emerges in which HF slopes are significantly increased for all but the ZT21-24 bin (Fig 7E, right panel), yet the most robust flattening of slopes clearly occurs at the beginning of lights-on. A two-way ANOVA reveals a 3-h bin × Treatment Day interaction (HF: F_{(3.745, 26.21)} = 5.667, p = 0.0024), with Dunnett’s post hoc tests demonstrating significant pairwise differences in 1/f slopes between BL_avg and W2 for each time point except ZT21-24. For REM, there is a significant steepening of lights on/off slopes (slopes become more negative) beginning at W3 for LF EEG signals (S9C Fig in S1 File, middle panel), but we noted that quantitatively, the effect sizes are quite small. NREM HF slopes recover to baseline values by W8 in the 12-hour bin analyses. However, the more granular 3h analysis still shows significant differences in NREM LF slopes on W8 from ZT3-9 (Fig 7D, right panel). This suggests that oxycodone withdrawal may still have transient long-lasting effects.

Chronic saline does not change sleep-wake rhythms.

For all rats, we used a within-subjects design in which telemetry data collected after cessation of drug infusions (i.e., withdrawal data) for each rat are compared to that rat’s average baseline data (BL_avg). However, this design does not allow direct detection of any deterioration of EEG/EMG signal over time or possible effects of rat age, weight, and general health on EEG/EMG signal or sleep stage architecture and dynamics. To control for these possibilities, 4 rats received 14 days of saline instead of oxycodone, and sleep data from W2, W4, W6, and W8 were analyzed. Because there are only 4 saline rats (and only 3 used for EEG dynamics analyses), experimental power for quantitative analysis is low. However, qualitative plotting of 3-h bin saline data for % time in NREM, NREM relative power and NREM LF/HF 1/f slopes across B4-B6 and W2, W4, W6, and W8 demonstrates a clear maintenance of the magnitude and temporal rhythmicity of sleep architecture and dynamics throughout the experiment (S2 Fig in S1 File).

Together, these findings under control conditions indicate that the telemetry devices maintain reliable and consistent EEG/EMG recordings for at least the length of our experiments, and that the observed sleep disruptions measured during W2-W8 of spontaneous oxycodone withdrawal can be specifically attributed to oxycodone withdrawal itself.