Increased heroin intake and relapse vulnerability in intermittent relative to continuous self‐administration: Sex differences in rats

Studies using intermittent‐access drug self‐administration show increased motivation to take and seek cocaine and fentanyl, relative to continuous access. In this study, we examined the effects of intermittent‐ and continuous‐access self‐administration on heroin intake, patterns of self‐administration and cue‐induced heroin‐seeking, after forced or voluntary abstinence, in male and female rats. We also modelled brain levels of heroin and its active metabolites.

. Unfortunately, many decades of preclinical research have not led to significant progress in the treatment of this condition. There is now a growing understanding that more refined animal models are required to understand the neurobiological basis of OUD and to develop more effective therapeutic approaches (Kreek et al., 2019;Venniro et al., 2020). The present study was designed to help address this issue.
One of the distinctive clinical features of OUD is an intense drug craving, which can be elicited by drug-paired cues and consequently lead to relapse into drug use (O'Brien et al., 1992;Preston et al., 2018). In 1986, based on clinical observations, it was suggested that craving progressively increases during abstinence (Gawin & Kleber, 1986). A similar phenomenon, called 'incubation of craving' (Grimm et al., 2001), was observed in rats after forced abstinence from several drugs, including heroin (Shalev et al., 2001;. In these previous studies, incubation of craving was characterized using a continuous-access drug self-administration. In this procedure, the drug is continuously available (except for a 20-s timeout imposed after each injection) and most rats develop a typical drug-taking pattern. The initial period of the self-administration session is characterized by a 'loading' phase featured by a high rate of infusions, which likely reflects the animal attempt to increase the brain-drug concentration above a 'satiety threshold' (Ahmed & Koob, 1998;Ahmed et al., 2000;Tsibulsky & Norman, 1999). After this loading phase, single infusions spaced apart are typically observed. This pattern of responding is often referred to as the 'maintenance' phase, during which animals titrate brain-drug concentrations to a steady state (Tsibulsky & Norman, 1999;Zimmer et al., 2011). Nevertheless, this self-administration procedure does not resemble the typical pattern observed in people with heroin use disorder.
Heroin abusers usually take large doses of heroin, interspersed by long abstinence periods (many hours or days) (Dole et al., 1966;McAuliffe & Gordon, 1974;Mello & Mendelson, 1987;Ross et al., 2008). As a consequence, drug brain levels are not maintained but decline overtime (Allain et al., 2015). This pattern can be modelled in rats through the intermittent-access self-administration procedure, originally developed by Zimmer and colleagues (Zimmer et al., 2012). In this seminal study, during the 6-h daily sessions, cocaine was available in 12 epochs of 5-min-binge periods-separated by 25-min periods during which the drug was not available-abstinence periods. Under these conditions, rats showed a pattern of cocaine self-administration characterized by closely spaced infusions (bursts) that would result in spiking, rather than steady cocaine brain concentrations (observed under continuous-access conditions) (Zimmer et al., 2012).
Studies that have directly compared continuous and intermittent access to cocaine in rats have shown that repeated spikes in the estimated cocaine brain levels are associated with an increased motivation to take and seek drugs (Algallal et al., 2020;Calipari et al., 2013;Fragale et al., 2021;James et al., 2019;Zimmer et al., 2012), particularly in females (Algallal et al., 2020;. Furthermore, an intermittent access to cocaine potentiates incubation of craving in both sexes after different periods of forced abstinence (James et al., 2019;Nicolas et al., 2019). Notably, cocaine craving was exacerbated in female rats during the oestrus phase of the What is already known • Preclinical research on opioid use disorder (OUD) is stymied by a translational problem.
• Intermittent, relative to continuous, access produces stronger motivation to seek and take cocaine and fentanyl.

What this study adds
• Intermittent access generates a stronger motivation for heroin and reveals sex differences in cue-induced craving.

What is the clinical significance
• The intermittent access results in drug-taking patterns that closely resemble the behaviour of heroin users.
• The refinement of animal models of OUD allows for translation of novel medications.
oestrous cycle (Nicolas et al., 2019). No such comparative studies have been conducted with heroin self-administration, which has been investigated using almost exclusively continuous-access procedures. Given the well-known differences between psychostimulant and opioid use disorder, at both clinical and preclinical levels (Badiani et al., 2011), it is important to fill this gap in the literature.
Thus, in the present study, we first compared continuous-and intermittent-access heroin self-administration in male and female rats and modelled the brain concentration of heroin and its active metabolites, 6-monoacetylmorphine (6-MAM) and morphine (Inturrisi et al., 1983), based on available drug kinetics data (Gottas et al., 2013). We then assessed the impact of these distinct selfadministration procedures on relapse to drug-seeking after two different abstinence conditions, forced and choice-based voluntary abstinence Venniro et al., 2020;Venniro et al., 2021; Azrin, 1976). This is exemplified in contingency management therapy (Epstein & Preston, 2003), where nondrug rewards (e.g. monetary vouchers), given in exchange for being drug-free, can maintain abstinence for many months. However, when contingency management discontinues, most individuals relapse (Roll, 2007;Silverman et al., 2012). Thus, we recently introduced a choice-based voluntary abstinence to overcome another limitation of the early incubation studies, where a forced abstinence period was experimentally imposed Venniro et al., 2016).

| Experimental overview
The goal of the experiment was to compare the effect of two different drug self-administration procedures, continuous and intermittent access on (1) drug-taking and related self-administration patterns, (2) modelled brain concentrations of heroin and its active metabolites, (3) incubation of heroin craving, after forced or voluntary abstinence and (4) sex differences and the role of the ovarian hormones on craving. The experiment consisted of three phases: self-administration training, an abstinence period (either forced or voluntary), and relapse tests. After the relapse tests, the oestrous cycle was measured in the female rats. The investigators were blind to the training conditions ( Figure 1).

| Self-administration apparatus
Rats were trained in self-administration chambers located inside sound-attenuating cubicles, fitted with an electric fan and controlled by a custom-made system. Each chamber was equipped with a stainless-steel grid floor and two operant panels placed on the left and right walls (see Figure S1). The left panel of the chamber was equipped with a house light that signalled the insertion and subsequent availability of the heroin-paired active (retractable) lever.
Responses on this lever activated the infusion pump and the discrete white-light cue located above the lever and heroin was delivered through a modified cannula (Plastics One; Roanoke, VA, USA) connected to a liquid swivel (Instech; Plymouth Meeting, PA, USA) via F I G U R E 1 Timeline of the experiment polyethylene-50 tubing that was protected by a metal spring. In addition, the left wall was equipped with an inactive (stationary) lever that had no reinforced consequences. The right panel was equipped with the palatable solution-paired active (retractable) lever. Responses on this lever activated the infusion pump and the three-light cue located above the lever. The 1-ml palatable solution was delivered to a receptacle located near the solution-paired lever, connected with a silicon tubing to a syringe that contained the palatable solution.

| Palatable solution self-administration
The training procedure was similar to the one described in previous studies Caprioli et al., 2017;Rossi et al., 2020;

| Intravenous surgery
Rats underwent intravenous catheterization after palatable solution self-administration. Rats were anaesthetised with isoflurane (5% induction, 2-3% maintenance) (Alcyon Italia S.p.A., Roma, IT) and injected with carprofen (2 mgÁkg À1 , subcutaneous injection; Zoetis Italia S.R.L., Roma, IT), immediately after the surgery and for the following 5 days to relieve pain and decrease inflammation. The silastic catheter was inserted into the jugular vein as previously described

| Heroin self-administration
Heroin self-administration training was divided into two phases, acquisition and training. In the acquisition phase, rats were trained to selfadminister heroin (0.1 ml per 3-s; 0.075 mgÁkg À1 per infusion) 2 hÁday À1 for 3 days (maximum infusions = 20) on a fixed ratio 1 (FR1), 20-s timeout reinforcement schedule. The sessions started with the insertion of the two levers (active and inactive) and the illumination of the house light. Responses on the active lever (fixed ratio 1) were reinforced by heroin infusions, paired with the cue light, followed by a 20-s timeout during which lever pressing was not reinforced and the cue light was on. Subsequently, for the training phase, rats were divided into two groups (matched for their total heroin intake during acquisition, Table S2): One group was trained using the continuousaccess procedure (n = 26 male, n = 37 female) and the other using the intermittent-access procedure (n = 27 male, n = 34 female). In the continuous-access condition, the rats had continuous access to heroin 6 hÁday À1 on fixed ratio 1, 20-s timeout reinforcement schedule. The sessions started with the insertion of the two levers and the illumination of the house light. Responses on the active lever were reinforced by heroin infusions, paired with the cue light, followed by a 20-s timeout during which lever pressing was not reinforced and the cue light was on. In the intermittent-access condition, the rats had access to heroin for a total of 60-min during the 6-h daily session, in twelve 5-min ON periods (drug available) and 25-min OFF periods (drug unavailable and levers retracted) under fixed ratio 1 reinforcement schedule with no timeout. The length of the OFF periods was set based on a previous study (O'Neal et al., 2020) and corresponded to the time needed for heroin and 6-monoacetylmorphine to dissipate (Gottas et al., 2013). Each 5-min ON period started with the insertion of the two levers and illumination of the house light and ended with the retraction of the levers and shutdown of the house light. Responses on the active lever were reinforced by heroin infusions, paired with a cue light (3-s), followed by no timeout. The rats were trained in continuous-or intermittent-access conditions for 10 days (training phase; maximum infusions = 90 per day to prevent overdoses) and after every three consecutive drug selfadministration sessions, the preference between palatable solution and heroin was assessed with discrete choice tests (see below).

| Discrete choice tests
The discrete choice sessions were conducted using the same parameters (dose of heroin and dose of SM7% solution per reward and stimuli associated with the two active retractable levers) used during the training phase. Rats were allowed to choose between the heroinpaired and the SM7% solution-paired levers in a discrete trial choice procedure. Each 160-min choice session was divided into 20 discrete trials that were separated by 6-min. Briefly, each trial began with the presentation of the house-light followed 10-s later by the insertion of both the SM7% solution-paired and heroin-paired levers. Rats then had to select one of the two levers. The operant response requirement for the lever's selection was set to two consecutive responses (FR2) to avoid accidental choices (Vandaele et al., 2016). If the rats responded within 2-min, they received the reward corresponding with the selected lever. Reward delivery was signalled by the heroinassociated or SM7% solution-associated cue (20-s), the retraction of both levers and the house-light turning off. If the rats failed to respond on either active lever within 2-min, both levers were retracted and the house-light was turned off with no reward delivery.

| Modelling brain levels of heroin and its metabolites
The theoretical brain levels of heroin and its active metabolites, 6-MAM and morphine were estimated in representative rats using the  (Table S1). In this study, the authors assessed levels of heroin and its metabolites, in blood and brain extracellular fluid, in male Sprague-Dawley rats, after a single passive i.v. heroin administration (1.3 mg). For a detailed account of the pharmacokinetic modelling, refer to the Material S1.
One methodological consideration in the present study derives from the fact that brain concentrations were modelled from chronic i.v. heroin self-administration data based on parameters derived from an acute i.v. administration in male rats (Gottas et al., 2013). While the present pharmacokinetic findings (Figure 4) should therefore be interpreted with caution, it is unlikely that this aspect would have significantly altered the modelled brain levels. This is because previous clinical studies based on intravenous, oral and subcutaneous administration of opioids (oxycodone, codeine and morphine) did not report significant changes in acute vs chronic opioid metabolism (Kimbrough et al., 2020;Zernig et al., 2007). Finally, brain levels of heroin and its metabolites were modelled only in male rats, since two studies revealed sex differences in opioids pharmacokinetics after intraperitoneal or oral administration (Chan et al., 2008;Djurendic-Brenesel et al., 2010). Notably, in the present study, no sex differences in heroin self-administration patterns were observed, including differences in total intake, frequency of intake, inter-infusions intervals ( Figure 3).
This suggests that the pharmacokinetic determinants of the frequency and pattern of intravenous heroin self-administration across sexes are not dissimilar.

| Abstinence period
After the relapse test on day 1, continuous and intermittent rats were randomly assigned to the forced or voluntary abstinence condition.

| Voluntary abstinence
Rats (continuous access n = 9 male/n = 13 female, intermittent access n = 12 male/n = 14 female) were allowed to choose between heroin (one infusion) and palatable solution (one delivery) during 20 discrete-choice trials for 18 days .

| Forced abstinence
Rats (continuous access n = 12 male/n = 14 female, intermittent access n = 15 male/n = 13 female) were brought to their home cages and handled twice a week.

| Relapse tests
Rats were tested for heroin-seeking under extinction conditions on abstinence days 1 and 21 (after forced or voluntary abstinence).
The duration of the test sessions was 30-min to minimize the carryover effect of extinction learning on day 1, which may decrease drug-seeking on day 21 . The sessions began with the illumination of the house-light, followed 10-s later by the insertion of the heroin-paired lever; the house-light remained on for the duration of the session. Lever presses during the tests resulted in the contingent presentation of the light cue, previously paired with heroin infusions, but no heroin infusion was delivered.

| Oestrous cycle
In female rats (n = 48) the oestrous cycle was monitored daily (for 5 days) before and immediately after the relapse test on day 1 and day 21 by a vaginal cytological test (Nicolas et al., 2019). A cotton tip, moistened with saline, was rolled into the vaginal orifice to collect the vaginal smear (Goldman et al., 2007). Then the cotton tip was rolled onto a microscope slide and analysed within 5 min, using an Olympus BX51 microscope (20Â magnification; RRID:

| Statistical analysis
The manuscript complies with the recommendations of the British Journal of Pharmacology on experimental design and analysis (Curtis et al., 2018). The statistical analysis was undertaken only for studies where each group size was at least n = 5. The sample size and animal numbers were determined by power analysis of pre-existing data Caprioli et al., 2017;Fragale et al., 2021;Nicolas et al., 2019;. The group sizes are the number of independent observations and the statistical analyses were performed using these independent observations. The group size for female rats was much larger than the group size for male rats since female rats, after the relapse tests, For the palatable solution self-administration, the data were analysed separately for rewards using the between-subject factor of Sex (male, female) and the within-subject factor of Session. For heroin self-administration, data were analysed separately for infusions using the between-subject factors of Sex and Access (continuous, intermittent) and the within-subjects factor of Session. For the discrete choice tests and voluntary abstinence, we normalized the indifference level between palatable food pellets and heroin (preference score) at 0 using the following formula: [1 À (% drug choices/50%)] (Lenoir et al., 2007) and analysed the data using the between-subject factors of Sex and Access and the within-subjects factor Session. For the continuousaccess group, the average of the inter-infusion intervals in the 10th self-administration session was analysed using the between-subject factor of Sex. For the intermittent-access group, the average of infusion during the 5-min ON periods in the 10th self-administration session was analysed using the between-subject factor of Sex. For the cumulative infusions in the 10th self-administration session, the cumulative 1-min infusions were analysed using a multifactorial analysis and the between-subject factor of Sex. For the minute x minute infusions, in the intermittent-access group, the average of infusions earned during each minute was analysed using the between-subject factor of Sex and the within-subject factor of Session. For the relapse tests, the active lever presses were analysed using the between-subjects factors of Sex, Access and Abstinence condition (forced, voluntary) and the withinsubjects factor of Abstinence Day (1, 21) and we included the inactive lever presses as a covariate. For the oestrous cycle in the relapse test day 1, the active lever presses on day 1 were analysed using the between-subjects factors of Access and Cycle phase (oestrus, nonoestrus). In all analyses of the relapse tests, the number of inactive lever presses was used as a covariate to statistically control for the effect of abstinence period on non-specific (training independent) lever presses during testing.

| RESULTS
The multifactorial ANOVAs yielded multiple main and interaction effects; thus, only significant effects, which are critical for data interpretation, are reported in this section (see Table S2 for a complete reporting of the statistical analyses and their exact P values).
3.1 | Sucrose + maltodextrin 7% (SM7%) selfadministration During acquisition, male and female rats increased their SM7% intake and lever pressing over sessions ( Figure S2A,B, left). Female rats selfadministered significantly more mlÁkg À1 of SM7% relative to their body weight compared to male rats ( Figure S2A, left). During training, female rats had a significantly higher SM7% intake relative to male rats ( Figure S2A, right).

| Heroin self-administration
During acquisition, no sex differences were observed in the number and frequency of infusions (minute x minute infusions) and active and inactive lever presses (Figure 2a-c and Table S2). During training, male and female rats increased their heroin intake over time in both access conditions, but the total heroin and frequency of intake were significantly higher in the intermittent-access relative to the continuous-access condition (Figure 2a,b). Notably, the intermittentaccess rats, both male and female, earned most of the infusions during the first minute of the 5-min ON periods (Figure 3b).
Additionally, a progressive increase in heroin intake during the first minute was evident across sessions ( Figure 3b). Furthermore, the temporal pattern of self-administration in either the intermittent-or the continuous-access conditions did not differ between sexes ( Figure 3a and Table S2). Finally, the oestrous cycle did not influence heroin self-administration regardless of the access conditions ( Figure S3). Notably, the lack of sex differences in our study agrees

| Voluntary abstinence
During the voluntary abstinence, rats from both access conditions displayed a strong preference for SM7% relative to heroin and increased choices for SM7% across sessions ( Figure 5).

| Incubation of heroin craving
Incubation of heroin craving was observed in the continuous-but not in the intermittent-access condition after forced abstinence, with higher lever pressing on abstinence day 21 than on day 1. In particular, rats in the intermittent-access condition did not show an 'incubated' profile because of the higher heroin-seeking on abstinence day 1, which was similar to day 21 (Figure 6a, left and centre). The time course of heroin-seeking on abstinence day 1 was significantly higher in the intermittent-relative to the continuous-access condition (Figure 6b, left). No differences between access conditions were observed on abstinence day 21 (Figure 6c, left). In the continuousaccess condition, voluntary abstinence prevented incubation of heroin craving (Figure 6a, centre), as previously reported . In the intermittent-access condition, lever responding on both abstinence day 1 and day 21 was higher in female rats than in male rats; this effect was independent of the abstinence condition ( Figure 6a, right).

| Incubation of heroin craving and the oestrous cycle
We observed no significant effect of the oestrous cycle in the incubation of heroin craving (Figure 7a). Female rats, in both access conditions, displayed a similar lever pressing during the oestrus and nonoestrus phase during the relapse tests.

| DISCUSSION
There are three main findings in our study. First, the overall heroin intake and modelled peaks brain levels of heroin and 6-MAM were much higher during the intermittent than the continuous-access selfadministration. Second, intermittent access to heroin was followed by higher heroin-seeking during early abstinence (day 1) and that remained stable over time (day 21). This phenomenon was more pronounced in female rats. Third, the oestrous cycle was not associated with the magnitude of relapse to heroin-seeking, regardless of training conditions.

| Heroin intake and modelled brain concentrations of heroin and its active metabolites during intermittent-and continuous-access self-administration
The main unexpected finding was that despite the much shorter drug access, heroin intake was significantly higher in the intermittent-than in the continuous-access condition (Figure 2a). This higher intake was accompanied by a self-administration pattern characterized by closely spaced infusions (bursts) mainly concentrated in the first minute of access ( Figure 3b). In contrast, the continuous-access condition was featured by a more regular pattern of intake, single infusions spaced apart (Figure 3a, left). In the following paragraphs, we discuss two distinct, but not mutually exclusive explanations, accounting for these divergent results.
The divergent results across access conditions might therefore be due to the long timeout imposed after each injection (20 s in our study, but can vary up to 40-s in other studies) present in the continuous-but not in the intermittent-access condition.
Heroin has a very short terminal half-life ($0.9-min in the rat brain; Gottas et al., 2013). Therefore, in the continuous-access condition, during the timeout, carboxylesterases are already metabolizing heroin to 6-MAM (Andersen et al., 2009;Gottas et al., 2014), a phenomenon that counteracts heroin accumulation in the brain. This is exemplified by the low heroin peaks observed in the continuous-relative to the intermittent-access condition (Figure 4b). On the other hand, the lack of this timeout period in the intermittent-access condition allows rats to infuse heroin in a burst-like pattern. This pattern of self-administration leads to a significant heroin accumulation in the brain (Figure 4b). In other words, the rats trained under continuousaccess conditions cannot reach the high peaks observed in the intermittent-access condition, no matter the rate of lever pressing.  (Gottas et al., 2013). In this condition, the levels of 6-MAM fluctuate around a steady state.
According to the hypothesis highlighted above, rats in the intermittent-access condition are more likely to experience the rewarding and intoxicating effects of high levels of heroin. In contrast, rats in the continuous-access condition, because of the timeout, are unable to achieve the same high brain levels of heroin and as a consequence, they self-administer heroin to maximize the rewarding effect of its metabolite 6-MAM.
Notably, in both access conditions, brain concentrations of morphine increased throughout the session and decreased slightly between injections. Indeed, the intervals between infusions were 10-15 min, which is shorter than the half-life of morphine ($50-min) (Gottas et al., 2013) (see Figures 3a and 4a). This finding suggests that morphine does not contribute significantly to the differences observed between heroin access conditions or the temporal dynamics during the self-administration, a phenomenon that most of the historical literature would not have predicted. Indeed, morphine is still widely believed to be the main, if not the only, metabolite responsible for the effects of heroin (Oldendorf et al., 1972).
Notably, although the metabolic breakdown of heroin is well known, the pharmacological activity of its active metabolites and their association to the acute and chronic effects are still surprisingly overlooked and will require further research.
The second explanation resides in the learning determinants of drug-taking. Previous studies, in rats trained under continuous-access conditions, have suggested that animals self-administer drug to achieve a specific drug 'satiety threshold' and learn to titrate drug levels around this threshold (self-administering drug injections whenever drug levels in the body drop below the threshold) (Panlilio et al., 2003). Thus, rats in the intermittent-access condition could have learned to self-administer several injections to maintain drug levels in the body above the drug satiety threshold during the 25-min OFF period. While this is a plausible explanation that warrants further studies, it should be noted that studies directly comparing continuous and intermittent access to cocaine showed that drug intake was significantly lower in the intermittent-access relative to continuousaccess condition (Algallal et al., 2020;James et al., 2019;Nicolas et al., 2019;Zimmer et al., 2012), a finding directly in opposition to our results. Notably, a similar dissociation in drug intake between psychostimulants and opioids has been previously reported by Panlilio and colleagues (Panlilio et al., 2003) in rats trained to self-administer cocaine or remifentanil, a short-acting opioid with a t 1/2 $ 0.7-min (Haidar et al., 1997).
Finally, in agreement with the preclinical literature on heroin self-administration (Bossert et al., 2021;Stewart et al., 1996;, we did not observe sex differences in the total intake (Figure 2a) or the pattern of heroin selfadministration in either the intermittent-or continuous-access conditions (Figure 3a,b). These results match the clinical evidence reporting no sex differences in the amount of heroin consumed (Gjersing & Bretteville-Jensen, 2014;Kennedy et al., 2013), the number of injections during drug-taking periods (Ross et al., 2008) or in the degree of enjoyment of use (Kennedy et al., 2013). Our  Institutes of Health (NIH) mandate of including both males and females in preclinical biomedical studies. Therefore, we tentatively suggest that in studies explicitly addressing the neurobiological determinant of heroin self-administration, it may not be necessary to use sex as an independent variable and double the n per experimental condition . Instead, we suggest including both males and females (preferably equal number) in each experimental condition, as advocated by Joel and McCarthy (Joel & McCarthy, 2017) for behavioural models in which sex differences are not observed. However, our data are at odds, with an earlier study reporting higher heroin intake in female compared to the male rats (Lynch & Carroll, 1999). This discrepancy, as previously F I G U R E 6 Relapse tests. (a) Relapse (incubation) test. Each left side of the graphs shows the data from the forced abstinence condition, while the right side from the voluntary abstinence condition. Data are mean ± SEM of lever presses on the active lever during the 30-min extinction test on abstinence day 1 and on day 21. * Different from continuous access, P < 0.05. # Different from abstinence day 1, P < 0.05. § Different from males, P < 0.05. (b) Time course relapse test abstinence day 1. Data are mean ± SEM of lever presses at each 10 min of the test session on day 1. * Different from continuous access. (c) Time course relapse test abstinence day 21. Data are mean ± SEM of lever presses at each 10 min of the test session on day 21 (voluntary abstinence: continuous access males n = 9/females n = 13, intermittent access males n = 12/females n = 14; forced abstinence: continuous access males n = 12/females n = 14, intermittent access males n = 15/females n = 13) discussed , is likely the result of the significantly lower unit dose (0.015 mgÁkg À1 ), a fifth of the unit dose used in the present study.

| Heroin relapse
A second relevant finding in our study is that rats in the intermittentaccess condition after a forced abstinence did not show 'incubated' drug craving, a phenomenon that has been consistently observed after forced abstinence from continuous access to heroin Shalev et al., 2001;Theberge et al., 2012; and both continuous and intermittent access to cocaine (Gueye et al., 2019;Nicolas et al., 2019). The lack of 'incubated' heroin craving observed in the intermittent-access condition is likely due to the high heroin-seeking observed in the early phases of abstinence ( Figure 6a). This strikingly resembles the high levels of initial craving typically observed in treatment-refractory active users being prescribed heroin maintenance (Blanken et al., 2012), in heroin users seeking treatment at different time points during abstinence (Wang et al., 2012), in abstinent users on methadone substitution therapy (Blanken et al., 2012) and in users who have completed therapeutic (Childress et al., 1986).
From a mechanistic perspective, it is known that a number of factors, including sex, context and schedule of drug reinforcement affect the development and time-course of sensitization of the dopamine system (Lefevre et al., 2020;Robinson & Berridge, 1993;Stewart & Badiani, 1993;Vanderschuren et al., 1997). Indeed, several studies have shown that the high motivation for drug and drug-associated cues observed after an intermittent access to cocaine is mediated by a sensitized dopamine response, which is not present in rats trained under continuous-access conditions (Calipari et al., 2013;. This sensitized dopamine response resembles the enhanced dopamine response to drug and drug cues observed in clinical imaging studies (Jasinska et al., 2014;Samaha et al., 2021). Based on the rationale provided above, we speculate that the repeated bursts of high heroin concentrations produced by the intermittent access could induce an extremely rapid sensitization of relevant neural F I G U R E 7 Effect of oestrous cycle in incubation of craving after continuous or intermittent heroin self-administration in female rats. substrates, resulting in an intense cue-induced craving since the very early phases of abstinence. An alternative mechanistic explanation can be found in the reward allostatic hypothesis of substance use disorder (Koob & Le Moal, 2001). In the intermittent-access condition, the repeated abstinence periods (lasting 25-min), which divide the drug-taking periods of heroin (12 epochs), would repeatedly trigger compensatory mechanisms, down-regulating the 'reward system' and determining a persistent high heroin craving (Koob, 2020). Further preclinical and clinical studies will be required to fill this gap in the literature.
We also observed that in the intermittent-access condition, drugseeking was significantly higher in female rats relative to males ( Figure 6a). This is in agreement with clinical studies showing shorter abstinence periods and higher reactivity to heroin-associated cues in women, relative to men (Petry & Bickel, 2000;Yu et al., 2007). A 'telescoping effect' in the transition to substance use disorder and a high relapse rate in women has been suggested (Brady & Randall, 1999) to be driven by ovarian hormones (Becker & Chartoff, 2019). With psychostimulants, vulnerability to relapse is increased during the follicular/oestrus phase (Nicolas et al., 2019).
Nevertheless, evidence for the role of ovarian hormones in opioidseeking has not been established (Knouse & Briand, 2021). Thus, we have explored the role of the oestrous cycle in the incubation of heroin craving, after a prolonged forced or voluntary abstinence. Our results indicated that the oestrous cycle does not influence cueinduced heroin-seeking, with a similar lever pressing in the oestrus and non-oestrus phases of the cycle (Figure 7a). This is consistent with a recent study demonstrating that treatment with oestradiol or progesterone does not influence cue-induced reinstatement of heroin-seeking (Vazquez et al., 2020). The distinct neurobiological mechanisms involved in relapse to opioids and psychostimulants (Badiani et al., 2011) may help account for these results, with different classes of drugs interacting differently with ovarian hormones.
Finally, after both continuous and intermittent access to heroin, the food choice-based voluntary abstinence completely suppressed heroin taking, regardless of the training conditions ( Figure 5). In addition, in confirmation to our previous findings , we observed a lack of sex differences in voluntary abstinence ( Figure 5). These results are consistent with those of Reiner and colleagues (Reiner et al., 2020), showing no sex differences in voluntary abstinence from fentanyl and are consistent with the lack of sex differences in the efficacy of contingency management in promoting abstinence in humans (Epstein et al., 2009

| Concluding remarks and clinical implications
In 1987, Mello and Mendelson (1987) suggested that to improve treatments for substance use disorder, a comprehensive understanding of the factors regulating drug use is essential. One of those factors is the pattern of drug-taking (Allain et al., 2015). In our study, we compared continuous and intermittent access to heroin and found that the intermittent access resulted in higher peaks of heroin and 6-MAM. This pattern more closely resembles the behaviour of heroin users who generally take high doses of heroin through the most rapid route of administration (i.v. injection) to experience a euphoric 'rush' (McAuliffe & Gordon, 1974;Seecof & Tennant, 1986). In intermittentaccess rats, the fast-rising peaks of heroin and 6-MAM levels may indeed represent the pharmacokinetic correlate of this highly soughtafter subjective effect. Most habitual users re-use the drug not only to reproduce the desired effect but also to suppress withdrawal signs/ symptoms (Koob & Le Moal, 2001;Sinnett et al., 1980), which usually occur within 2-12 h of the last dose. In clinical practice, it provides the rationale for the most widely used pharmacotherapy for heroin dependence, substitution therapy with either methadone or buprenorphine (Gerra et al., 2003). These longer acting drugs produce stable opioid levels throughout the day, with well-documented benefits on heroin craving (Greenwald, 2002), withdrawal (Kleber, 2008) and other indirect drug-associated consequences (Strasser et al., 2010).
As discussed earlier, the utility of the intermittent-access procedure is also supported by the high heroin-seeking that remains stable from an early to a late phase of abstinence (particularly in females), while the continuous-access condition was followed by an increase in craving overtime. Consistent with this, a clinical study reported that individuals with OUD characterized by an occasional (intermittentlike), but 'intense', heroin use have a higher risk of relapse during the initial period of abstinence relative to more 'regular' users (Termorshuizen et al., 2005).
Together, these observations indicate that the intermittentaccess condition may reflect many features of heroin abuse in humans, including administration of large doses of heroin to reach intoxicating effects, high craving in the early abstinence from heroin use and higher relapse susceptibility in females. Thus, the intermittent-access procedure would be useful in investigating the mechanisms underlying the development of heroin use disorder and possibly new approaches for treatment.

AUTHOR CONTRIBUTIONS
DC, GD, IR and MM carried out the experiments and FB, GD and IR performed data analysis. DC, GD, IR and MV designed the study and wrote the manuscript with AB, DR and MSM. All authors critically reviewed the content and approved the final version before submission.

CONFLICT OF INTEREST
The authors declare that they do not have any conflicts of interest (financial or otherwise) related to the content of the paper.

RIGOUR
This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the British Journal of Pharmacology guidelines for Design and Analysis and Animal Experimentation, and as recommended by funding agencies, publishers and other organizations engaged with supporting research.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.