Female and male rats readily consume and prefer oxycodone to water in a chronic, continuous access, two-bottle oral voluntary paradigm

Abstract

The increasing abuse of opioids - such as oxycodone - poses major challenges for health and socioeconomic systems. Human prescription opioid abuse is marked by chronic, voluntary, oral intake and sex differences. To develop interventions, the field would benefit from a preclinical paradigm that similarly provides rodents with chronic, continuous, oral, voluntary and free-choice access to oxycodone. Here we show female and male rats voluntarily ingest and choose oxycodone over water and show both dependence and motivation to take oxycodone during a chronic oral voluntary, two-bottle choice, continuous access paradigm. Adult female and male Long-Evans rats were given unlimited, continuous homecage access to two bottles containing water (Control) or one bottle of water and one bottle of oxycodone dissolved in water (Experimental). Virtually all experimental rats voluntarily drank oxycodone (~10 mg/kg/day) and escalated their intake over 22 weeks. Females self-administered twice as much oxycodone by body weight (leading to higher blood levels of oxycodone) and engaged in more gnawing behavior of wooden blocks relative to males. Precipitated withdrawal revealed high levels of dependence in both sexes. Reflecting motivation to drink oxycodone, ascending concentrations of citric acid suppressed the intake of oxycodone (Experimental) and the intake of water (Control); however, Experimental rats returned to pre-citric acid preference levels whereas Controls rats did not. Pre-screening behaviors of rats on open field exploration predicted oxycodone intake. Thus, rats consumed and preferred oxycodone over time in this chronic two-bottle oral choice paradigm and both sexes displayed many features of human oxycodone abuse.

1. Introduction

Opioid misuse and abuse in the US and globally pose major challenges for health and socioeconomic systems worldwide (Cicero and Ellis, 2017a; Compton and Volkow, 2006; Zacny et al., 2003). Opioids such as oxycodone are mainstays of pain management (Cowan et al., 2003; Minhas and Leri, 2017). but the relatively easy access to these medications and their common diversion has led to a higher incidence of addiction in the exposed population (Cicero and Ellis, 2018, 2017a; 2017b). This increase has deadly consequences; in 2014, oxycodone ranked third behind heroin and cocaine in causing overdose deaths (Warner et al., 2016). As initial use of opioids can result in dependence that leads to concomitant or replacement use of more accessible, less expensive but more lethal opioids, i.e. heroin or fentanyl, it is critical to understand the factors underlying prescription opioid abuse liability (Cicero and Ellis, 2018).

To understand these factors, preclinical researchers would benefit from animal models of prescription opioid abuse liability that recapitulate key aspects of long-term use, such as voluntary and escalating intake that results in dependence. Prescribed oxycodone is typically taken orally, and animal models of oral opioid intake have been in use for many years (Cappell and LeBlanc, 1971; Enga et al., 2016; Heyne, 1996; Klein, 1995; Nichols et al., 1956; Risner and Khavari, 1973; Shaham et al., 1992; Stolerman and Kumar, 1972; Wikler et al., 1963; Yanaura et al., 1980). Some oral intake paradigms predispose the rodent to opioid drinking via intraperitoneal injection of opioids (Nichols, 1963; Seevers and Deneau, 1963), or force opioid drinking by restricting access to non-drugged water (Cappell and LeBlanc, 1971; Kumar et al., 1968; McMillan et al., 1976; Thompson and Ostlund, 1965). Although useful for assessing the impact of opioids on physiology and behavior, such paradigms confound the motivational component that characterizes voluntary drug-seeking behavior and opioid reinforcing properties (Meisch, 2001). One classic, true voluntary oral intake model is the “two-bottle choice” paradigm, widely-used in studies of rodent alcohol intake (Belknap et al., 1993; Meisch and Beardsley, 1975; Sinclair, 1976; Taylor et al., 1990; Weiss et al., 1990) but also employed for studies of rodent opiate intake (Hill and Powell, 1976). Whereas some two-bottle studies adulterate the opioid solution with an alternative reinforcer, such as sucrose (Alexander et al., 1981), others found rats would drink an unadulterated opioid solution and even prefer it to water (Heyne, 1996). The use of long-term, free access to an opioid solution in addition to water has provided useful insights, showing, for example, that during periods of controlled drug choice opioid intake can be modified by environmental and individual factors (Heyne, 1996; Pelloux et al., 2006).

Despite this progress, there remain three knowledge gaps in the current literature on long-term, continuous free access to opioids via two-bottle choice. First, morphine and other opioid-agonists have been assessed in two-bottle choice, but oxycodone has not. This is a notable gap in the literature given oxycodone’s distinct pharmacokinetic profile and its non-medical use worldwide. Second, although studies have examined oral opiate intake in the two-bottle choice paradigm, many have used acute, intermittent, or sub-chronic two-bottle access. Long-term continuous access studies can provide additional clinically-relevant insight into the acquisition, maintenance, and potential tolerance to oral opioids. Third, although pre-exposure endophenotypes can predict later opioid use (such as high vs. low locomotor activity in a novel environment (Flagel et al., 2014; Piazza and Le Moal, 1996)), such pre-screening has not been performed for oxycodone in a long-term voluntary choice paradigm. Although not a central focus of this current work, sex differences in human and rodent oral opioid intake have been long appreciated (Alexander et al., 1978; Craft, 2008; Fillingim and Gear, 2004; Graziani and Nisticò, 2016; Hadaway et al., 1979; Klein, 1995; reviewed in Serdarevic et al., 2017). However, female rodents have not been assessed in a paradigm that affords long-term, continuous access to oral oxycodone. Overall, the development of a long-term voluntary choice opioid oral intake model would complement other rodent models of dependence and addiction, such as intravenous (IV) self-administration, thereby facilitating intervention testing.

Here we address these knowledge gaps by employing a modification of a classic 2-bottle paradigm to provide rats with long-term, free choice, and continuous access to an oral oxycodone solution and water. Adult experimental female and male Long Evans hooded rats were given free access (24 h/d, 7 d/wk) to a bottle containing oxycodone and a bottle of plain water in their home cage (Experimental rats), whereas Control rats had access to two bottles of water. Virtually all Experimental rats spontaneously and largely drank oxycodone, escalating their intake over time. Female and male rats had similar intake of and preference for oxycodone, but by body weight females self-administered twice as much oxycodone with resultant higher blood levels. Females also engaged in more gnawing of wooden blocks and Nylabones relative to males. Motivation to self-administer oxycodone (via citric acid suppression test) revealed that both Experimental and Control rats suppressed their intake of citric acid-adulterated oxycodone or water, respectively. However, Experimental rat oxycodone intake immediately returned to pre-citric acid intake levels, whereas Control rat water intake did not. Oxycodone dependence via naloxone-precipitated withdrawal was evident in Experimental but not Control rats. Screening before oxycodone access on open field test predicted oxycodone ingestion in males and many of the behaviors on these tests correlated with long-term levels of intake. Overall, this model recapitulates many features of human oxycodone abuse in female and male rats, and therefore will be a useful addition to the field for probing the neural and physiological mechanisms that mediate long-term voluntary drug use and exploration of potential treatment options.

2. Material and methods

2.1. Animal welfare statement, animals, and husbandry

Animal procedures and husbandry were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and performed in IACUC-approved facilities at the Children’s Hospital of Philadelphia (CHOP) Research Institute (AAALAC Accreditation #000427, PHS Animal Welfare Assurance D16–00280 [OLAW A3442–01]). The study was planned to generate the greatest possible amount of information from the fewest number of animals. Female (n = 12) and male (n = 12) Long-Evans Hooded rats were purchased from Envigo (formerly Harlan) at the age of 60–65 days. Rats were individually-housed in a temperature-controlled (20–23 °C) and humidity-controlled (30–70%) housing room in an AAALAC-approved vivarium at the Colket Translational Research Building at the Children’s Hospital of Philadelphia (CHOP) Research Institute. Rats were housed in a ventilated cage (L: 50.8, W: 40.6, H: 20.3 cm) with ad libitum food and water available via lixits until after behavioral prescreening (see section 2.2). The lights were turned off at 6:00 PM and turned on at 6:00 AM. Rats were handled 3 times per week for 2–3 min (min) to familiarize them with different experimenters. Rat weights were recorded once before behavioral testing and once a week throughout the experiment until terminal procedures were performed. The cage was removed and cleaned once a week by experimenters (not by husbandry staff), with an occasional spot change for cages deemed unsanitary. Corn cob bedding was supplemented with paper shreddings and a red 3 × 6 inch polycarbonate tunnel (Animal Specialties and Provisions, cat# TPI001) that were changed and/or replenished during the weekly cage change. Cage location in the rack and in the vivarium room was maintained throughout the experiment to minimize disruptions.

2.2. Baseline behavioral assessment

To prescreen rats, after ~2 weeks of individual housing rats were tested in the open field (open field 1, OF1) and for marble burying behavior (Fig. 1). The data from these tests were used to 1) assure that the random assignment to subsequent Experimental and Control groups was unbiased for the traits measured; and 2) to determine if the behaviors from these tests could be used to predict later outcome measures once oxycodone was available. Open field was also repeated during Week 31 (open field 2, OF2) as a post-screening evaluation.

Fig. 1. Pre-screening behavioral assessment (before group assignment) revealed no overt group differences.

(A) EXperimental time course. Long Evans (LE) hooded rats, 12 females (F) and 12 males (M), were pre-screened (Weeks −3 and −2) on the open field (OF1) and marble burying (MB) tests. After pre-screening, all rats entered the Baseline (Water/Water) period where they were habituated for two weeks (Weeks −1 and 0) to drink from two water bottles, one each in front of or behind the food hopper positioned on the right side of the cage (see panel 1F). Starting at Week 1, Control rats (3 female and 3 male rats, randomly chosen) continued to receive two bottles containing water (Water/Water); Experimental rats (9 female and 9 male rats, also randomly chosen) received an Oxycodone-containing bottle in the front position and a water-containing bottle in the back position (Water/Oxycodone). At the end of Week 11/start of Week 12, the front bottle (containing water or oxycodone) was switched to the back position for both Control and Experimental rats (Bottle Position Switch). Tail vein blood was extracted from all rats, once during Week 15 and once during Week 21, both time performed ~9 h into the 12-hr dark cycle. From Weeks 27–29, all rats were intermittently exposed to increasing concentrations of citric acid (CA aversion) to test for aversion. Week 31, all rats were tested for nociception in the plantar thermal test (PTT) and open field again (OF2) and subsequently assessed for signs of naloxone precipitated withdrawal (PW).

(B) The analysis of total distance traveled over the 30 min in the OF1 via a two-way ANOVA (Sex x Treatment) revealed no main effect of Sex (p = 0.247), no main effect of Treatment (p = 0.314), and no interactions (p = 0.815). Note that in this panel B and panels C-E and G-I, Experimental and Control refers to the subsequent assignment of naive rats to these two groups.

(C) Analysis of number of entries into the Center area using a two-way ANOVA (Sex x Treatment) revealed no main effect of Sex (p = 0.215), no main effect of Treatment (p = 0.308), and no interaction (p = 0.425).

(D) Analysis of duration of time spent in the OF1 Intermediate area by a two-way ANOVA (Sex x Treatment) revealed a significant effect of Sex (p = 0.015), no main effect of Treatment (p = 0.806), and no interaction (p = 0.679).

(E) Marble burying test analysis of the number of marbles buried using a two-way ANOVA (Sex x Treatment) revealed no main effect of Sex (p = 0.735), no main effect of Treatment (p = 0.885), and no interaction (p = 0.363).

(F) Schematic of the home cage during the two-bottle habituation (Baseline) period, with the food hopper on the right side of the cage and water bottles placed in the front and the back of the hopper as depicted. Rats had access to both bottles 24 h/day and 7 days/week. Not drawn to scale.

(G) Analysis of total water intake during the Baseline period via two-way ANOVA (Sex x Treatment) revealed no main effect of Sex (p = 0.407), no main effect of Treatment (p = 0.507), and no interaction (p = 0.421).

(H) Analysis of bottle position preference (ratio of intake from the front bottle over the total intake) during the Baseline period via two-way ANOVA (Sex x Treatment) revealed no main effect of Sex (p = 0.282), and no main effect of Treatment (p = 0.443), and no interaction (p = 0.644).

(I) Analysis of body weight during the Baseline period via two-way ANOVA (Sex x Treatment) revealed a significant main effect of Sex (p < 0.001), no effect of Treatment (p = 0.797), and no interaction (p = 0.294).

The open field was performed as previously described (Bouwknecht et al., 2007; Kulesskaya and Voikar, 2014) with minor modifications. The open field was conducted in a square arena (L: 76, W: 76, H: 30 cm) with white plastic floor and white walls. The open field arena illumination ranged from 135 (corner) to 225 (center) lux with an average lux of 150. Open field was run between 1:00 PM and 5:00 PM. The rat was placed into the arena in the same position each time (e.g. bottom right corner). The following behavioral parameters were scored using ANY-maze software: total distance traveled, average speed, maximum speed in the Center and Intermediate zones of the open field, longest time period spent in the Center and Intermediate zones, number of entries into the Center and Intermediate zones, duration of time in the Center and Intermediate zones, average duration in the Center and Inter-mediate zones, latency to enter the Center and Intermediate zones, and number of fecal boluses deposited during test. For reporting purposes, a subset of the behavioral parameters are shown: total duration, entries in the Center zone, and duration in the Intermediate zone. One week later, rats were then assessed for marble burying, as previously described (Deacon, 2006; Ku et al., 2016; Njunge and Handley, 1991; Schneider and Popik, 2007; Thomas et al., 2009) with minor modifications. Briefly, rats were habituated for 10 min in a new housing cage bottom filled with approximately 5 cm corn cob bedding. The rat was removed and 18 marbles (1 cm diameter each) were placed in the cage in a 3 × 6 pattern. The rat was then reintroduced to the cage for 20 min period of testing and behavior was recorded (Handycam, HDR-CX330 9.2). Rats were tested in groups of 4 each, and 3 blinded observers each collected a score for each rat. A consensus was reached after visual evaluation of each cage with further validation from the offline video score as-needed.

2.3. Oxycodone self-administration

After behavioral assessment (Fig. 1), lixits were removed from all cages and rats were given water via a bottle (Hydropac^® or Zyfone bottle, as described below) for the duration of the experiment. Intake of liquid was measured at 7:00 AM and 4:00 PM each day until Week 11, and at only 7:00 AM from Week 11 until the end of the experiment. Liquid intake was calculated as the change in weight (g) of the bottle from the previous measurement. A ratio was calculated as the amount of liquid drunk (g) from one of the two liquid bottles over the total intake from the two bottles. During Baseline (Weeks −1 and 0, Fig. 1A), rats were provided two water Hydropacs (in front and back position) in a hopper on the right side of the cage with food placed in between. Starting on Week 1, Control (“Water/Water”) rats continued to receive water from both bottles, whereas Experimental (“Water/Oxycodone”) rats received water in one bottle and oxycodone dissolved in water in the other. The starting oxycodone concentration in each bottle was determined for each rat based on their baseline average water intake and their baseline body weight. This allowed the rats to titrate their own intake, and resulted in a delivered concentration ranging from 0.06 to 0.12 mg/mL, with a goal of individual rat intake of ~10 mg/kg/day (Schrott et al., 2008) if they exclusively drank from the oxycodone bottle. From Week 3–11, the oxycodone bottle was placed in the front hopper position, and from Week 12–32 it was switched to the back hopper position (see “Bottle position switch”, Fig. 1A). At Week 8, Hydropacs were replaced by water bottles (ZyfoneTM One CageTM, 16 oz) with a twist cap (ZyfoneTM high-temperature plastic One CapTM) and a 1.5″ Sipper Tube With Stainless Ball (Part#-ST3.0SB, Alternative Design Mfg & Supply Inc.) due to extensive chewing of the Hydropac plastic valve and spillage by some rats that made the intake data unreliable. Thus intake data from Week 3–8 were not analyzed and are not shown. One female Experimental rat was excluded from analyses because of extensive chewing and spillage of oxycodone from the Hydropac and bottle, making her data unreliable.

2.4. Home cage behaviors

Home cage behaviors were observed twice daily, at 7:00 AM and 4:00 PM. A single rat was tracked by one observer during 1-min observation periods for the presence or absence of specific behaviors (Table 1) indicative of normal activity (i.e. grooming, walking) and spontaneous withdrawal-like behavior (i.e. wet dog shakes) as adapted from previous studies (Gellert and Holtzman, 1978; Jones and Barr, 1995; Koob et al., 1992; Rasmussen et al., 1990).

Table 1.

Spontaneous and precipitated withdrawal behaviors.

Spontaneous and precipitated withdrawal behaviors
Behavior	Definition	How recorded?
Head movement	Lateral and rotary motion of the head	0 ≥ 1 with +1 if occurs > 1X
High walking	Elevated torso or spine curved upwards, can present as abnormal posture in rats.	0 ≥ 1 with +1 if occurs > 1X
Jumping	Sudden leaping such that all four paws are off the bottom of the chamber	0 ≥ 1 with +1 if occurs > 1X
Ptosis	One or both eyelids closed or semi-closed	0 = absent; 1 = mild; 2 = moderate; 3 = marked
Wet dog shakes	Rapid shaking of the whole body	0 ≥ 1 with +1 if occurs > 1X
Burrow	Sliding the head under the corn cob cage bedding	0 ≥ 1 with +1 if occurs > 1X
Diarrhea	Extremely loose and watery stool	0 = absent; 1 = mild; 2 = moderate; 3 = marked
Yawning	Wide opening and closing (stretching) of the mouth. Can be seen with and without body stretching (separate behavior).	0 ≥ 1 with +1 if occurs > 1X
Stretching	Extension or dorsal flexion of the trunk causing a visual lengthening of the body	0 ≥ 1 with +1 if occurs > 1X
Teeth chatter	Movement of the jaw with constant chattering of the teeth, or distinct repetitive movement of cheek/facial muscles.	0 ≥ 1 with +1 if occurs > 1X and each occurrence separated by 3 s

Only precipitated withdrawal behaviors
Behavior	Definition	How recorded?
Writhing	Abdominal stretching	0 ≥ 1 with +1 if occurs > 1X
Sweeping tail	Heavily carrying the tail around	0 ≥ 1 with +1 if occurs > 1X
Chewing	Without matter in the mouth	0 ≥ 1 with +1 if occurs > 1X
Ejaculation	Releasing semen	0 = absent; 1 = mild; 2 = moderate; 3 = marked
Piloerection	Bristling of hair	0 = absent; 1 = mild; 2 = moderate; 3 = marked
Straub tail	Curved upwards tail	0 ≥ 1 with +1 if occurs > 1X

2.5. Serum assessment of oxycodone levels

During Weeks 15 and 21, tail vein blood was extracted from all rats ~9 h into the 12-h dark cycle. Blood was taken via the tail vein under isoflurane anesthesia. Blood was collected in sterile Eppendorf tubes with 10 μl of heparin and centrifuged at 4,000 RPM for 10 min and serum separated and stored at −20 °C. Serum was sent to the CHOP Bioanalytical Core Center for Clinical Pharmacology. Assay development consisted of oxycodone separation through ultra-performance liquid chromatography (UPLC) and detection by tandem mass spectrometry (MS-MS) using a triple quadrupole mass spectrometer (API4000).

2.6. Gnawing home cage behavior

At the start of Week 12, each rat cage received 1 Nylabone (Nylabone DuraChew Wolf, Animal Specialties and Provisions cat# NW103) and 1 wood gnawing block (Bio-Serv, cat# K3512, medium size). The weights of the Nylabone and block were recorded daily at the same time of day. The change in weight of each was calculated to derive a gnawing index, defined as the total amount in grams of Nylabone and wood block chewed over time.

2.7. Citric acid (CA) aversion test

Between Week 27 and 29 all rats received ascending concentrations of citric acid (CA, Sigma Aldrich, cat# PHR1071) at 1, 3, and 5 mM in the back bottle. 5 mM CA was repeated once (referred to as 5 and 5’). In between each CA exposure, rats received their original bottle (water or oxycodone) for 2 days to allow for a return to baseline. Liquid intake, body weight, preference, and deviation from baseline were scored daily.

2.8. Plantar thermal test

On Week 31 all rats were tested for analgesia in the plantar thermal test using a Hargreaves apparatus (IITC Inc. Life Science, Model 390G, Series 8, Heated Base) as previously described (Barr et al., 2016). Rats were habituated for 10 min in a custom acrylic enclosure with high walls placed on the tempered (30 °C) glass surface of the apparatus. The test was performed on the plantar surface of each hindpaw by a focused, radiant heat light source creating a 4 × 6 mm focal spot set to a measured temperature of 60 °C; in preliminary work with other rats, these parameters typically produce a baseline hindpaw withdrawal latency of ~10–15 seconds (s). If the rats did not respond, a cut-off of 20 s was employed to prevent tissue damage. Each hindpaw was measured 3 times with approximately 1–2 min between tests.

2.9. Precipitated withdrawal

On Week 32, all rats were tested for signs of withdrawal. Rats were habituated for 10 min to a clean cage containing corn cob bedding. Behavior was recorded for entire session using ANY-maze software and 3 side cameras. After habituation, each rat was injected with naloxone (1 mg/kg, i.p. Sigma-Aldrich cat# PHR1802) to induce precipitated withdrawal. Withdrawal symptoms as previously described and summarized in Table 1 (Bläsig et al., 1973; Cicero et al., 2002a; Harvey-Lewis et al., 2015; Koob et al., 1992; Rasmussen et al., 1990) that occurred during the 15 min post-injection period were hand-scored by trained observers. Weight was measured before habituation and after 15 min of precipitated withdrawal. One weight data point for one animal was lost from the male Experimental group.

2.10. Statistical analysis

Data analysis was performed using GraphPad Prism^® (version 8; Macintosh) or PAST3 (version 3.2). A two-tailed unpaired t-test was used to test the difference of the means between two groups. A 2-way repeated measures (RM) ANOVA or a mixed-effects mode was used to test for main effects of Treatment or Sex and their interaction. If an interaction was significant after analysis of variance, a post-hoc test for multiple comparisons with correction was performed. Dunnett’s post-hoc test was performed when comparisons were made to a single control group; Sidak’s or Tukey’s tests were performed for multiple comparisons between different groups. Data are presented as individual data points and as mean ± SEM. In all analyses, the level of significance was p < 0.05 and 3 significant digits are provided. Complete statistical analysis output for every panel is provided in Supplemental Table 1 for main figures and Supplemental Table 2 for supplemental figures.

A multivariate sample-based factor analysis (PAST3; Q-Mode CAB-FAC (Imbrie and Kipp, 1971)) was conducted separately on the pre-screening open field and marble burying behaviors for females and males, obtaining a single score for each rat, by multiplying each pre-screening open field behavior measures by the factor score of that behavior. That summed score was then correlated with the dependent measures during the intake phase (precipitated withdrawal behaviors, spontaneous withdrawal behaviors, dose, intake, preference, CA, blood, gnawing). In addition, individual open field behaviors from both the pre- and post-screening open field tests were correlated with oxycodone dose at Week 22, when intake was stable.

3. Results

3.1. Pre-screening for a predictive endophenotype: assessment prior to experimental assignment revealed no overt group differences before group assignment

3.1.1. Pre-screening behavioral assessment

Before any manipulations (Week −3 and Week −2; Fig. 1A), rats were pre-screened for open field (OF1) behavior, as an indicator of both anxiety and locomotor activity (Colman, 2001; Markel’ et al., 1989) and for marble burying, as a repetitive behavior (Thomas et al., 2009) often associated with an anxiety-like phenotype (Njunge and Handley, 1991). For OF1, the total distance traveled (Fig. 1B) and the number of entries into the Center zone (Fig. 1C) were similar within groups irrespective of Sex or the to-be-assigned Treatment (Control or Experimental). There was a significant main effect of Sex of duration in the Intermediate zone (Fig. 1D), with females spending ~3x more time in the Intermediate zone than males. There was no difference in the number of marbles buried driven by Sex or to-be-assigned Treatment groups (Fig. 1E). Thereby at pre-screening, the behavior parameters were similar, with the exception that females showed higher levels of novelty-seeking (longer time spent in the Intermediate zone in the OF1 relative to males).

3.1.2. Baseline water/water intake

At Baseline (average of Week −1 and Week 0, Fig. 1A), lixits were removed from cages and two bottles of water were provided to both Control and Experimental rats in their home cage. This allowed the rats to habituate to the bottles and the “free-choice” paradigm (freedom to drink water from either front or back bottle; Fig. 1F). There was no effect of Treatment or Sex in the average amount of liquid ingested over the first two weeks of habituation (Fig. 1G). There was no effect of Treatment or Sex in the preference for the front or the back bottle (Fig. 1H). Within Sex, rats had similar body weight (Fig. 1I), with females weighing - as expected - less than males. Thus, the random assignment to either Control or Experimental group resulted in unbiased inclusion in the two groups.

3.2. Experimental rats increased their oxycodone preference and drank more oxycodone over time, with females ingesting more than males and having higher resultant oxycodone blood levels

3.2.1. Two bottle-choice period

During the 2 bottle-choice period which began Week 1 (Fig. 1A), female Experimental rats increased their total liquid intake on Weeks 18–22 compared to Week 1, whereas male Experimental rats drank the same total intake throughout (Fig. 2A). For oxycodone intake, female Experimental rats ingested more oxycodone from Week 9 onwards compared to female Week 1, whereas male Experimental rats ingested more oxycodone from Week 13 onwards compared to male Week 1 (Fig. 2B). In addition, when the bottle position was switched (Week 11), males increased their intake on Weeks 14–22 relative to Week 11 (Fig. 2B). In contrast, Controls had a stable intake of water (Fig. 2C) and preference for the bottle position (Fig. 2D). In regard to preference for the oxycodone-containing bottle, both Experimental females and males increased their preference for oxycodone by Week 9 relative to Week 1 (Supplemental Fig. 1). When the bottle position was switched, both females and males maintained their preference for oxycodone, but male rats subsequently increased their preference for oxycodone by Week 17 compared to the week before the bottle switch (Week 11; Supplemental Fig. 1).

Fig. 2. Experimental rats increase liquid intake and oxycodone preference over time in a sex-dependent manner.

(A) Total liquid intake in Experimental rats (receiving Water/Oxycodone beginning Week 3) analyzed via a restricted mixed-effects model (Sex x Time) revealed no effect of Sex (p = 0.393), a significant effect of Time (p < 0.001), and a significant interaction (p < 0.001). Post-hoc analyses within each Sex showed that compared to Week 1, males did not change their liquid intake over time, whereas females did from Week 18 onwards (denoted by an * for each week; *p < 0.05, **p < 0.01, ***p < 0.001). Post-hoc analysis of total intake between females and males showed no sex difference for any given week.

(B) Oxycodone intake in milliliters by Experimental rats analyzed via a restricted mixed-effects model (Sex x Time) revealed no effect of Sex (p = 0.247), a significant effect of Time (p < 0.001), and a significant interaction (p = 0.019). Post-hoc comparison within Sex showed that both females increased their intake starting at Week 9 relative to Week 1, whereas males increased their intake relative to their Week 1 value starting at Week 13 (Females: Week 1 vs. Weeks 9–22 p < 0.001; Males: Week 1 vs Week 13–22 p ≤ 0.001; lines above Weeks 9–22 and 13–22 denote different from Week 1 of that Sex). When the bottle position was switched, post-hoc analyses revealed females did not change their intake of oxycodone (Week 11 vs. Weeks 12–22 did not differ), but male rats escalated their intake of oxycodone by Week 14 and maintained that increased intake through Week 22 (* denotes male value different from male Week 11; *p < 0.05, **p < 0.01, ***p < 0.001).

(C) Total liquid intake in Control rats (receiving Water/Water throughout) analyzed via two-way RM ANOVA (Sex x Time) revealed no effect of Sex (p = 0.086), no effect of Time (p = 0.086), but a significant interaction (p = 0.006). Post-hoc analyses within each Sex showed that compared to Week 1, neither females nor males differed over time in their liquid intake. Post-hoc analysis of total intake between females and males showed no difference at any given week.

(D) Preference for front bottle in Control rats (receiving Water/Water throughout) analyzed via two-way RM ANOVA (Sex x Time) revealed no effect of Sex (p = 0.962), no effect of Time, (p = 0.323), and no interaction, (p = 0.690).

3.2.2. Body weight over time

The body weights of female Control and female Experimental rats did not differ from each other at any time point (Fig. 3A). Male Experimental rats weighed significantly less than Control male rats from Week 13 onwards (Fig. 3A). To assess oxycodone amount self-administered by the female and male Experimental rats, we calculated both the absolute intake (mg, Fig. 3B) and intake by body weight (mg/kg, Fig. 3C) for each rat. Female and male rats self-administered higher absolute amounts of oxycodone over time, achieving a high steady state intake from Week 9–22 relative to Week 1 (Fig. 3B). Females self-administered higher dose of oxycodone per body weight on Weeks 9 through 22 relative to Week 1, whereas males self-administered higher dose per body weight on Weeks 15 through 22 relative to Week 1 (Fig. 3C). In addition, females ingested greater mg/kg oxycodone relative to males on Weeks 9–12, 17–18, and 20–22 (Fig. 3C). The greater oxycodone intake in females was reflected in higher oxycodone blood levels compared to males at both Week 15 (Fig. 3D, left panel) and Week 21 (Supplemental Fig. 2). When the dose of oxycodone was correlated with Week 15 blood levels, the relationship was significant and positive for males, but not significant for females (Fig. 3D, right panel), and the difference between the two correlations was significant.

Fig. 3. Experimental rats intake increasing amounts of oxycodone with resultant high blood levels of the drug, especially in females.

(A) Body weights over time in Control and Experimental rats analyzed via three-way ANOVA (Sex x Time x Treatment) revealed a significant effect of Time (p < 0.001), a significant effect of Sex (p < 0.001), but no effect of Treatment (p = 0.550). The three-way interaction (Sex x Treatment x Time) was significant (p < 0.001), and posthoc tests showed that female Control rats did not differ from female Experimental rats at any time point. Male Experimental rats weighed significantly less than Male Control rats from Week 13 onwards (* denotes difference in male Control and Experimental value at that week).

(B) Analysis of self-administered oxycodone intake (mg) via two way ANOVA (Sex x Time) revealed a significant effect of Time (p < 0.001), but no effect of Sex (p = 0.504), and no interaction (p = 0.065). Posthoc analyses of the main effect of time showed that both sexes differed on Weeks 9–22 compared to Week 1, but that there were no differences from Weeks 9–22 compared to Week 11.

(C) Analysis of self-administered dose via two-way ANOVA (Sex x Time) revealed a significant effect of Sex (p < 0.001), and a significant effect of Time (p < 0.001), but no significant interaction (p > 0.050). Post-hoc analysis of Time within same sex group showed that female Experimental rats self-administered a higher dose of oxycodone Weeks 9–22 compared to Week 1, whereas male Experimental rats self-administered oxycodone a higher dose Weeks 14–22 compared to Week 1 (* denotes within sex difference from Week 1, * p < 0.05, ** p < 0.01, *** p < 0.001). Post-hoc analysis of Sex at specific timepoints showed significant differences at Weeks 9–12, 17–18, and 20–22 (^∧ denotes female different from male at that timepoint; ^∧ p < 0.05, ^∧∧ p < 0.01, ^∧∧∧ p < 0.001).

(D) Left panel: Analysis via two-tailed, unpaired t-test of oxycodone blood levels between female and male rats drawn on Week 15 revealed a significant difference of Sex (p = 0.009). Right panel: Correlation analysis of the blood levels with oxycodone intake 12 h before blood collection revealed a significant positive correlation in males (r = 0.957; r² = 0.916, p < 0.001) but not for females (r = 0.128; r² = 0.016, p = 0.762). The difference between the two correlations was significant (p = 0.008).

3.3. Experimental rats show higher levels of dependency in precipitated withdrawal and increased gnawing behavior but no changes in nociception

3.3.1. Withdrawal in the home cage

Daily assessment of home cage spontaneous withdrawal symptoms (burrow, teeth chatter, stretch, and jump) at 7AM showed no significant difference between groups when analyzed as individual behaviors or when these behaviors were summed over 22 weeks (Fig. 4A). Overall, Experimental rats chewed more Nylabone and wooden block in the home cage between Weeks 12 and 22 of oxycodone exposure than did Control rats; moreover, Experimental females gnawed more than Control females, Control males, and Experimental males. Experimental males did not differ from Control males (Fig. 4B).

Fig. 4. Precipitated withdrawal and gnawing behavior are increased in experimental rats - particularly in females - relative to control rats.

(A) Individual spontaneous withdrawal signs analyzed via a two-way ANOVA (Sex x Treatment) showed no main effects of Treatment or Sex and no interaction for frequently occurring behaviors: Burrow (Sex, p = 0.230; Treatment, p = 0.081; interaction, p = 0.998), Teeth Chatter (Sex, p = 0.825; Treatment, p = 0.920; interaction, p = 0.229), Stretch (Sex, p = 0.637; Treatment, p = 0.721; interaction, p = 0.060), Jump (Sex, p = 0.707; Treatment, p = 0.091; interaction, p = 0.974). A two-way ANOVA of the total spontaneous withdrawal signs also showed no effect of Sex (p = 0.470) or Treatment (p = 0.071) and no interaction (p = 0.700).

(B) The sum of total amount in grams of chewed nylabone and wooden blocks were averaged over 11 weeks (Weeks 12–22). Two-way ANOVA (Sex x Treatment) revealed an effect of Sex (p = 0.030), an effect of Treatment (p = 0.001), and a significant interaction (p = 0.036). Post-hoc analysis within and between Sex revealed Experimental females gnawed more than Control females (p = 0.003), Control males (p = 0.002), and Experimental males (p = 0.002).

(C) Individual precipitated withdrawal signs analysis via two-way ANOVA (Sex x Treatment) revealed no effect of Sex, a significant effect of Treatment, but no interaction for the following behaviors: Burrow (Sex, p = 0.193; Treatment, p = 0.016; interaction, p = 0.620); Teeth Chatter (Sex, p = 0.731; Treatment, 0.005; interaction, p = 0.964); High Walk (Sex, p = 0.196; Treatment, p = 0.018; interaction, p = 0.196); Diarrhea (Sex, p = 0.303; Treatment, p = 0.002; interaction, p = 0.303). There was no effect of Sex or Treatment and no interaction for Stretch (Sex, p = 0.067; Treatment, p = 0.383; interaction, p = 0.503). Analysis of the total withdrawal signs via two-way ANOVA showed no effect of Sex (p = 0.278), a significant effect of Treatment (p < 0.001), and no interaction (p = 0.680). P-values are provided for each withdrawal measure on the X-axis.

(D) Weight loss after precipitated withdrawal analyzed via two-way ANOVA (Sex x Treatment) revealed no Sex effect (p = 0.897), a significant effect of Treatment (p = 0.044), but no interaction (p = 0.453).

3.3.2. Plantar thermal test

When measuring paw withdrawal latency in the plantar thermal test, there was no effect of Treatment nor Sex and no Treatment by Sex interaction. The means and SEM (in s) for the four groups were Control female: 13.93 ± 2.29; Experimental female: 12.53 ± 1.9; Experimental male: 11.14 ± 1.14; Control male: 14.74 ± 1.75. Details of the latencies for each trial are shown in Supplemental Fig. 3A. Moreover, the correlations between oxycodone intake averaged over the 4 days before the plantar test were non-significant for females, males, or the two sexes combined (r’s = 0.019; 0.356; 0.267 respectively; combined data for females and males: Supplemental Fig. 3B).

3.3.3. Open field

For OF2, most measures were similar within groups irrespective of Sex or Treatment (Control or Experimental), including total distance traveled and duration in the Intermediate zone (Supplemental Fig. 4). There was a main effect of Treatment for number of entries into the Center zone, with Experimental rats having more Center entries than Control rats.

3.3.4. Precipitated withdrawal

In contrast to home cage spontaneous withdrawal behaviors, Experimental rats presented more precipitated withdrawal symptoms than Control rats at Week 31 for the following discrete behaviors analyzed separately: burrow, teeth chatter, high walk, and diarrhea. When the total precipitated withdrawal score was analyzed, Experimental females and males presented more withdrawal signs than their respective controls (Fig. 4C). After precipitated withdrawal, Experimental rats lost more weight than did Controls: Experimental females lost four-fold more weight than Control females and Experimental males lost two-fold more than Control males (Fig. 4D).

3.4. Increasing concentrations of CA suppressed intake in both Control and Experimental rats, but Experimental rats returned to initial baseline preference whereas Control rats did not

When CA was added to the back bottle (Fig. 5A), all rats suppressed their intake from that bottle at higher CA concentrations compared to their intake the day before the first CA test (BL0; Fig. 5B). Experimental males suppressed at 3 mM and higher CA, whereas the other groups first suppressed at 5 mM (Fig. 5B). When CA was removed during the interspersed baselines, Experimental rats returned to the initial baseline preference, whereas Control rats maintained their aversion (Fig. 5C).

Fig. 5. Citric acid suppressed drinking in all rats, although experimental - and not control - rats returned to baseline preference.

(A) Schematic of home cage layout during the citric acid (CA) test. All rats had an unadulterated water bottle in the front bottle position (Water). On the other side of the food hopper, in the back bottle position, Control rats (left schematic) had a water bottle with increasing concentrations of CA (Water + CA) and Experimental rats (right schematic) had an oxycodone bottle (Oxy) with increasing concentrations of CA (every third day; 1, 3, and 5 mM; Oxy + CA). The 5 mM concentration of CA was repeated twice (termed 5 and 5′). CA days were interspersed with “baseline” days (BL, shown as 0 CA concentration in B and C).

(B) Preference for the back bottle of Control and Experimental rats analyzed via two-way RM ANOVA (CA concentration x Treatment) revealed a significant effect of CA concentration (p < 0.001), a significant effect of Treatment (p = 0.012), and a significant interaction (p = 0.043). Post-hoc analysis of the preference change over increasing concentrations of CA revealed Experimental males, but not Experimental females, Control females, nor Control males, suppressed at 3 mM CA (denoted by ^∧; p = 0.028) compared to the initial preference when there was no citric acid (CA concentration = 0; BLo). All rats significantly suppressed intake at both presentations of the 5 mM concentration (denoted by *): Experimental females (5 nM, p = 0.002; 5′ nM, p < 0.001); Control females (5 nM, p = 0.004; 5′ nM, p = 0.002); Control males (5 nM, p < 0.001; 5′ nM, p = 0.004); Experimental males (5 nM and 5′ nM, both p < 0.001).

(C) Analysis of change from initial baseline (BL0) via a two-way RM ANOVA (Return to BL0 x Treatment) revealed an effect of Return to BL0, p < 0.001, no effect of Treatment, p = 0.182, and a significant interaction, p < 0.001. Post-hoc analyses showed that Control female and Control male rats maintained their aversion compared to BL0 intake (Control females BL5, * p = 0.015; BL5′, ** p = 0.005; Control males BL5′, *** p < 0.001), whereas Experimental female and male rats returned to BL0 levels (Experimental females BL5, p = 0.788, BL5′, p = 0.741; Experimental males BL5, p = 0.640; BL5′, p = 0.907).

3.5. Pre-screening of open field behavior predicted oxycodone intake-related outcomes in both female and male rats

To provide an unsupervised picture of the relationship of open field behavior to dependent measures during the oxycodone intake phase (intake, preference, blood levels, precipitated withdrawal, spontaneous withdrawal, CA, gnawing, etc.), we conducted a factor analysis of all pre-screening open field behaviors and marble burying. The analysis revealed a single factor for both females and males that explained ~71% of the variance each. The scatterplots for these analyses are shown separately for females (Supplemental Fig. 5A) and males (Supplemental Fig. 5B). To obtain a single score for each rat, the pre-screening open field behavior measures were multiplied by the factor score of that behavior and summed. Each rat’s factor score was then correlated with their dependent measures during the oxycodone intake phase. For females, no behaviors significantly correlated with open field measures. For males, significant correlations were seen with oxycodone dose at Weeks 9 and 11 (Fig. 6A).

Fig. 6. Factor analysis and correlations among pre-screening behaviors and post-oxycodone outcomes.

(A) Heat map of the predictions from the pre-screening open field test behaviors to oxycodone outcome variables. The factor loadings for each behavior for each sex were determined by a factor analysis (scatterplots for each sex are shown in Supplemental Fig. 5). Only the first factor explained significant variance and was used for the correlations shown in the heat map. For each behavior, the score of that behavior (frequency, duration, or latency) was multiplied by its loading on Factor 1. These values were summed over all behaviors to provide a single overall pre-test score for each rat. This heat-map presents the multiple correlations (+1.0 to −1.0) between that Factor score and the post-oxycodone variables related to intake and withdrawal. For females, no prescreening behaviors significantly correlated with the pre-test score (all r < 0.707; n = 8). For males, the dose of oxycodone ingested in Weeks 9 and 11 was significantly correlated with the pre-test factor score (r > 0.666; n = 9; white asterisk).

(B) Heat map of the multiple correlations between the self-administered dose at Week 22 and the pre- and post-screening open field variables in both females and males. The significant p-value is highlighted by a white asterisk is (r > 0.707, n = 8 for females; r > 0.666 for males, n = 9). For both females and males, there were a number of significant correlations between the pre-screening open field behaviors and the oxycodone intake (dose) at Week 22. There were no significant correlations between post-screening behaviors and the dose at Week 22.

In addition to the factor analysis described above, analysis of the correlation data between pre-screening open field behavior and oxycodone intake at Week 22 revealed a positive relationship between activity measurements in the open field - particularly increased activity in the Center and Intermediate areas - and intake (Fig. 6B). In contrast, there were no significant correlations between oxycodone intake at Week 22 and post-screening open field (OF2) behavior (Fig. 6B).

4. Discussion

Important aspects of a preclinical model for human prescription opioid abuse liability include prolonged access to drug, voluntary and often increasing self-administration when a choice is provided, high levels of intake, drug dependence, and subsequent somatic signs of opioid withdrawal. Here we show that when provided to rats via a classic voluntary drug intake model (2-bottle choice), oxycodone recapitulates these key features of human prescription opioid abuse. Rats given chronic, continuous choice between oxycodone and water reliably demonstrate voluntary and escalating intake of oxycodone, measurable levels of dependence, and motivation to take the drug. We believe this preclinical model of prescription opioid abuse - which relies on oxycodone oral self-administration - is of significant translational relevance for understanding the mechanisms that underlie opioid addiction and ultimately will contribute to the ongoing research efforts to combat the growing opioid epidemic.

4.1. Conceptual and methodological pros and cons of long-term, voluntary, oral oxycodone intake via 2-bottle choice

The face validity of opioid self-administration models is based on an animal’s motivation to seek and take drugs (Spanagel, 2017) and to do so despite adverse consequences (Koob and Le Moal, 2001). Employing a classic 2-bottle choice paradigm, rats readily drank oxycodone and showed a strong preference for the oxycodone-containing bottle relative to the water bottle. Experimental rats escalated their oral oxycodone intake over time, suggesting it was reinforcing, and rebounded to these high levels of intake following citric acid suppression. Thus, rats in a chronic, continuous oxycodone 2-bottle choice model show motivation to drink in the face of adverse consequences.

Other opioid self-administration paradigms have relied on either oral intermittent access (OIA) orIV self-administration (Carroll and Meisch, 1984; Enga et al., 2016; Gellert and Holtzman, 1978; Grim et al., 2018; Heyne, 1996; Jimenez et al., 2017; Leung et al., 1986; Meisch, 2001; Nguyen et al., 2018; Nichols and Hsiao, 1967; Shaham et al., 1993; You et al., 2017). Both OIA and IV self-administration models can lead to dependence and tolerance, and IV models can demonstrate drug-seeking motivation. OIA models mimic a common route of administration of humans (oral), whereas the rapid onset of IV drug effects enables assessment of contingent cues and behaviors and mimics IV drug intake. However, these models have limitations. High first-pass metabolism of drugs in OIA opioid models results in a short bioavailability and longer latencies to induce the “high” compared to IV models (Chan et al., 2008). IV self-administration models often rely on short-term access, which confounds the assessment of drug effects vs. those caused by daily withdrawal (Badiani et al., 2011; Meisch, 2001; Weise-Kelly and Siegel, 2001). We believe the oxycodone 2-bottle choice model used here addresses some of these issues. Its advantages include: a) the translational relevance for orally-delivered drugs, b) voluntary intake and choice between natural rewards (food/water) and oxycodone, c) the chronicity of drug available 24 h/7d over months, d) the non-invasive methodological approach, and e) the absence of confounders, such as adulteration of oxycodone solution to encourage oral intake or stress of the animal to induce intake.

In comparison to the multiple studies in which opiates were administered orally but intermittently (Davis et al., 2010; Enga et al., 2016; Fan et al., 2018; Gellert and Holtzman, 1978; Meisch, 2001), there are fewer studies that employed long-term continuous oral intake of morphine or other opioids. Those more chronic studies started with low opioid concentrations that were increased by the experimenter (Badawy et al., 1982), mixed with a sucrose solution (Fuentes et al., 1978; Leung et al., 1986), or provided as the sole source of liquid (Gellert and Holtzman, 1978; Shaham et al., 1992), with ingestion often limited to a few weeks. To our knowledge, no studies to date allowed the rodent to freely self-titrate opioid intake over several months without some form of taste enhancement or liquid constraint. In the present study, female and male rats preferred and drank increasing amounts of oxycodone over weeks and months; in prior shorter duration studies, rats did not prefer the opioid (typically drinking < 50% of total liquid intake) and did not escalate their intake (Badawy et al., 1982; Gellert and Holtzman, 1978), with the exception of the opioid etonitazene (McMillan et al., 1976). In all studies of oral opioid intake, including the present one, when tested, rodents displayed withdrawal behaviors after challenge with an opioid antagonist, highlighting the ability of many different oral paradigms to result in opioid dependence.

One caveat to the present study is that the 2-bottle paradigm used here necessitated that the rats be individually housed for ~10 months, and thus the results should be interpreted in the context of this social isolation. Social isolation augments many effects of opioids including oral self-administration and locomotor activity (Alexander et al., 1981; Coudereau et al., 1996; Deroche et al., 1994; Francès et al., 2000; Hadaway et al., 1979; Sudakov et al., 2003), but reduces morphine-induced place preference and analgesia (Bozarth et al., 1989; Coudereau et al., 1997; Schenk et al., 1983; Wongwitdecha and Marsden, 1996). Future work could test whether the escalating levels of oxycodone drinking we find in both females and males would also occur in socially-housed animals, and could also assess whether the sex differences in intake are robust (given that we have relatively small within sex control subject number).

4.2. Oxycodone 2-bottle choice highlights sex differences in oxycodone intake and blood levels

Prescription opioid abuse liability is increasingly understood as being sex-dependent (Becker and Koob, 2016; Bobzean et al., 2014), which is characterized by the more rapid increase in the rate of prescription opioid abuse in women than men (Warner et al., 2016). Previous work suggests a role for sex hormones in opioid dependence and opioid reinforcing properties (Alexander et al., 1978; Bobzean et al., 2014; Cicero et al., 2003, 2002a; 2002b; Hadaway et al., 1979; Harte-Hargrove et al., 2015; Serdarevic et al., 2017), particularly in oral self-administration protocols (Stolerman and Kumar, 1972). Sex differences are also apparent in the pharmacokinetic, metabolic, and analgesic effects of oxycodone (Chan et al., 2008; Holtman and Wala, 2006; Neelakantan et al., 2015), and subtle differences have been observed in the liability for opioid abuse (Collins et al., 2016; Mavrikaki et al., 2017). Therefore animal studies show that, relative to males, females self-administer more opioids, are more vulnerable to their reinforcing effects, and become more physically dependent, (Alexander et al., 1978; Boyer et al., 1998; Cicero et al., 2003; Graziani and Nisticò, 2016; Hadaway et al., 1979; Lofwall et al., 2012; Lynch, 2018; Lynch et al., 2002; Serdarevic et al., 2017). In our work presented here, when rats were given a choice between a water bottle and an oxycodone-containing bottle, both sexes readily drank oxycodone and escalated their intake. However, females self-administered twice as much oxycodone by body weight as did males, with a resultant five-fold increase in blood levels, and also gnawed more than males. The preference for oxycodone shown by females peaked early and remained at high levels throughout, whereas this peak in preference appeared later for males. Moreover, in the oxycodone 2-bottle choice paradigm presented here, individual differences during pre-screening were predictive of intake and self-administered dose at Weeks 9 and 11 for males but not for females. Whereas these data suggest that the chronic, continuous oxycodone 2-bottle paradigm used here is suited to interrogate sex differences under a choice self-administration paradigm, our relatively low within-sex control subject numbers encourage additional study of sex-specific differences with this model using a more robust experimental design.

Blood levels of oxycodone were consistently higher in females than in males, likely reflecting their greater intake and dose. It is notable that male (but not female) blood levels were correlated with oxycodone dose at the earlier blood draw time. In humans, females have higher CYP3A4 activity suggesting a faster oxycodone elimination than males, with females displaying more vulnerability to oxycodone’s effects (Andreassen et al., 2011; Graziani and Nisticò, 2016). Although we did not investigate sex differences in pharmacokinetics, the greater dose self-administered by females vs. males in our paradigm could have been driven by sex differences in oxycodone metabolism. When oxycodone was administered by gavage to Sprague Dawley rats, females showed higher oxycodone blood levels than males, providing evidence for differences in oxycodone metabolism (Chan et al., 2008). However, other work found subtle differences in IV self-administration but no difference in oxycodone blood and brain levels, suggesting no pharmacokinetic differences (Mavrikaki et al., 2017), and that sex difference in oxycodone metabolism may be specific to the oral route of administration. It is tempting to speculate that the higher total liquid intake in females in the data presented here was due to a higher intake needed to keep oxycodone blood levels constant. This would argue in favor of a faster elimination rate (but see Chan et al., 2008), lower oxycodone sensitivity, and/or different first pass metabolism, resulting in greater intake at the same concentration compared to males. Further studies on blood sampling at earlier and multiple time points along with metabolite measures are warranted to test this hypothesis.

4.3. Oxycodone two-bottle choice as a model of dependence and stereotyped gnawing behavior

When the long-term free-access oxycodone intake models are used, most rats will escalate their drug intake over time, which contrasts with the lack of escalation in many short-term access paradigms (Badiani et al., 2011; Meisch, 2001). In addition, presumably chronic, continuous access allows rats to titrate their intake to avoid withdrawal. In support of this, spontaneous withdrawal signs were not statistically different between control and oxycodone rats, suggesting that spontaneous withdrawal in our long-term access paradigm is negligible. There are caveats though; our sampling for spontaneous withdrawal was time- and behavior-limited, and a more thorough analysis of spontaneous withdrawal signs in the oxycodone 2-bottle choice model is warranted. In contrast to the lack of spontaneous withdrawal, both female and male Experimental rats showed robust precipitated withdrawal signs (Bläsig et al., 1973; Gellert and Holtzman, 1978; Leung et al., 1986; Sudakov, 1991), with ~4-fold more naloxone-induced signs of physical dependence relative to Control rats. Thus, the oxycodone 2-bottle choice model will be useful for assessing opioid dependence and the efficacy of treatments targeted to relieve withdrawal.

Despite no significant differences in spontaneous withdrawal, in the oxycodone two-bottle choice model used here Experimental (but not Control) rats developed stereotyped gnawing behavior. Gnawing and other oral stereotyped behaviors can be induced by repeated high doses of morphine (Wennemer and Kornetsky, 1999) or stress and are mediated by endogenous opioids (Bergmann et al., 1974; Ernst and Smelik, 1966; Morelli et al., 1989). One source of stress in the two-bottle choice model is social isolation which reduces tolerance to morphine but also precipitated withdrawal (Broseta et al., 2005; Coudereau et al., 1996; Jiménez and Fuentes, 1993). Here rats were singly-housed for 10 months, longer than in prior work; however, isolation is unlikely the sole contributor to increased gnawing behavior, as no Control rat measurably gnawed. It is possible that the combination of oxycodone and isolation increased the likelihood of gnawing, and we might expect that group-housed oxycodone rats would gnaw less than isolated oxycodone rats. Because social isolation is more stressful for females than males (Beery and Kaufer, 2015; Hatch et al., 1965; Westenbroek et al., 2005), the increased gnawing in particular by Experimental females may be explained by an opioid-mediated activation of the dopaminergic mesolimbic system known to induce gnawing behavior (Bergmann et al., 1974; Morelli et al., 1989).

4.4. Oxycodone two-bottle choice: appetitive and reinforcing behavior, escalation, and incentive salience

When challenged to drink an aversive tastant added to the oxycodone bottle, most rats (Experimental and Control) suppressed their drinking from the citric acid-adulterated bottle. However, some Experimental rats maintained their preference for oxycodone despite the addition of citric acid. Both female and male Experimental rats readily returned to baseline levels when the unadulterated oxycodone bottle was returned, whereas Control rats maintained their acquired aversion. These data suggest that two-bottle choice of oxycodone has high face validity for modeling prescription opioid abuse liability despite adverse conditions and it provides for a broader use of this model to study motivational aspects of prescription opioid abuse, such as individual differences in the development of physical dependence, oxycodone reinforcement, and its motivational salience.

4.5. Pre-screening for predictive factors

Not all humans become addicted to drugs and those that do show different degrees of dependence (Brady and Randall, 1999; Cicero and Ellis, 2017b). We believe the individual differences that emerge during the oxycodone two-bottle choice model can be harnessed to identify individual vulnerabilities that may be predictive of oxycodone abuse liability. Previous work defining predictive factors that contribute to an addictive phenotype (Belin et al., 2016) shows that locomotor and exploratory activity and anxiety-like phenotypes can be predictive of later drug-seeking behavior (Markel’ et al., 1989; Piazza et al., 1989). In our present work, a composite of pre-screened open field behavior predicted intake at Weeks 9 and 11 in males but not females, and fewer open field behaviors positively correlated with the dose ingested later in the experiment at Week 22 for females than for males. A caveat of this conclusion is that females had consistently higher levels of oxycodone intake than males. Thus, the fewer predicted intake-specific variables for females compared to males may reflect a ceiling effect specifically for intake variables. This ceiling effect is less likely to explain the lack of correlation for non-intake-related variables, such as weight loss after precipitated withdrawal.

4.6. Test of nociception

We found no difference in the baseline latencies to withdrawal the hindpaw in a test of thermal nociception following months of oral oxycodone intake and demonstrable levels of oxycodone in blood, suggesting that either the levels in blood were not high enough to be analgesic or that the Experimental rats were tolerant to any analgesic effects. There are caveats to this conclusion. First, we only used a single measure of nociception, and other noxious stimuli (mechanical, inflammatory) might prove more relevant. Second, nociception was assessed only at a single time point following months of exposure and we did not follow the time course of oxycodone’s effects. Third, we did not test dose-response analgesic effects; even though baseline latencies were not different, chronic oxycodone might still shift the dose-response curve to acute opioid-induced analgesia. Given the controversies over the effectiveness and risk of long-term opioid treatment for chronic non-cancer pain, more comprehensive testing of opioid-induced analgesia in this model is warranted.

4.7. Conclusions

Relying on a large literature using two-bottle choice models of drug intake (Hill and Powell, 1976), we show that rats voluntarily drink increasing amounts of oxycodone when given continuous access and prefer it to water when given a choice between oxycodone and water. In addition to mimicking oral ingestion of oxycodone in humans, the two-bottle choice model we used here presents several advantages as an additional model to study the abuse aspects of this drug, including the provision of choice between water and drug and unlimited access to the drug, which allows the animal to titrate its oxycodone levels in contrast to experimenter-controlled dosing. When allowed to freely-drink oxycodone, both females and males increase their intake over time, show physical dependence as indicated by precipitated withdrawal and stereotyped gnawing behavior. Oxycodone-drinking Experimental rats rebound from an aversive citric acid challenge, whereas Control rats do not. Females ingested more oxycodone and had higher blood levels than males, setting the stage for the study of sex differences in the aspects of oxycodone intake and potentially its consequences.

Future work would address why and how individual rats differ in their intake, pre-existing risk factors that might predict high levels of intake, mechanisms underlying the sex differences, and the role of isolation in these studies. We believe that the oxycodone two-bottle choice model is germane to probing treatment options, pharmacological and environmental, and in particular the role of stress in the motivation to take opioids. Also, as there are differences in brain structure activation in oxycodone exposed verse naive rats given acute oxycodone (Iriah et al., 2019), imaging rats pre-and post-oxycodone in females and males during the course of their voluntary intake could provide insights into neurocircuitry of oxycodone use in the two sexes.

HIGH LIGHTS.

• Adult rats offered continuous choice of oral oxycodone vs. water preferred oxycodone.

• Rats self-titrated oxycodone intake, yet females ingested more oxycodone than males.

• Both sexes were motivated to drink oxycodone, as shown by a citric acid aversion test.

• Both sexes became dependent on oxycodone, as shown by precipitated withdrawal.