Introduction

Alcohol use disorder (AUD) represents a substantial financial and health burden on society. While men tend to drink more than women and are diagnosed with AUD at a higher rate, women are closing the gap1,2,3. Additionally, women develop alcohol misuse more quickly from casual use4 and are more vulnerable to negative physical and mental health consequences of drinking5,6,7. Beyond the progression and consequences of the disease, gender differences are also reported in drinking motives, with suggestions that females often drink more for reasons of coping while males drink out of impulsivity and positively reinforcing mechanisms8,9,10,11. Such differences in drinking motivations may reasonably reflect underlying differences in approach-avoidance behavioral tendencies12,13 in the face of competing motivations (conflict), with drinking females exhibiting more avoidant traits and drinking males displaying stronger approach bias. Indeed, there is evidence that female drinking is related to negative affect more than male drinking14,15,16,17, together suggesting that gender-specific approach-avoidance biases may underlie distinct patterns of drinking behavior, potentially influencing susceptibility to addiction.

Studies examining the relationship between approach-avoidance conflict and substance use have typically focused on whether drug administration biases rodents towards conflict approach or risk taking18,19,20,21,22,23,24, without fully exploring the bidirectionality of this relationship, which is readily possible with the use of rodent models25. The few studies that have investigated how conflict approach or risk taking predict future ethanol or cocaine self-administration reveal that this relationship is sex-dependent. Recently, work from our laboratory reported that greater conflict-avoidance in female rats was related to higher levels of voluntary ethanol drinking (in homecage) while there was no relationship between conflict approach-avoidance bias and ethanol drinking in males26. Other rodent work has demonstrated that risky decision-making positively predicts future cocaine self-administration in male rats27, although this may be dependent on both sex and duration of access, as shorter access to cocaine gives rise to risk-averse females (but not males) self-administering more cocaine28. At the same time, riskier choice predicted higher escalation of cocaine intake in both males and females across extended access28. However, it is of note that existing preclinical studies typically asses drinking using a singular dimension of intake (i.e. drinking level), and thus fail to capture the multidimensional pattern of aberrant drinking and alcohol-seeking observed in AUD and as defined by DSM-5, which includes a loss of control over intake, continued drinking despite negative consequences, inability to cut back on alcohol seeking, and persistent cravings11,29,30,31,32,33,34,35. There is evidence suggesting that the two constructs of compulsive vs. excessive drinking are dissociable in rodents32,36,37,38 and that there may be sex/individual differences in the manifestation of compulsivity, with female rodents often displaying greater compulsive ethanol drinking than males, in parallel with increased consumption of ethanol32,38,39,40,41 (but see Ref.42). Additionally, frontloading – an ethanol intake pattern in which drinking is skewed towards the onset of access – typically results in binge-like behavior43,44 and is thought to represent a valid measure of motivation44. This drinking pattern mirrors bingeing in humans45, which is associated with more disordered alcohol use7,46. Sex differences in frontloading preclinically, however, are rarely reported and remain understudied44. Thus, in studies relating drug use to approach-avoidance conflict behavior, it is important both to consider patterns of drinking beyond simple measures of intake and to explore possible sex differences.

The goal of the current study, therefore, was to assess the predictive power of approach-avoidance conflict behavior and frontloading on future drinking outcomes. To this end, we first assessed animals’ approach-avoidance bias under cued conflict. We then used a three-phase operant self-administration task in which the response requirement for ethanol reinforcement was increased (from Fixed Ratio (FR) 1- > FR3), and the session length shortened from 30 to 15 min, to induce binge drinking in the animals. Crucially, we recorded not only the total ethanol intake, but also the pattern of drinking in these operant sessions, to derive two distinct measures of frontloading: bingeing, or intake over a fixed time period early on in access, and onset skew, the concentration of total ethanol-seeking behavior occurring within an early time period. Finally, we assessed three criteria previously proposed for the evaluation of addiction vulnerability29: motivation for ethanol (measured in a progressive ratio task), persistence of ethanol seeking despite its unavailability (responding in extinction), and ethanol taking in the face of mild footshock punishment (compulsive drinking). Based on our previous work26, we hypothesized that approach-avoidance conflict behavior would relate to drinking patterns in a sex dependent manner, with conflict approach predicting more addiction-like phenotype in males than in females. Additionally, we hypothesized that frontloading would predict addiction-like drinking, with a high binge, low onset skew profile yielding the most vulnerable phenotype.

Methods

Subjects

56 adult Long-Evans rats (28 male, 28 female, Charles River, QC, Canada), arriving at approximately 225 g (males) or 150 g (females), were used in this study. A minimum sample size of 50 was considered necessary to run the desired multiple regression analyses. All data were intended to be included in the analysis, with an appreciation of individual differences, though two males were excluded for failure to acquire operant responding for ethanol, and one female was euthanized during the experiment. Therefore, the final sample of rats completing this experiment was 53 (26 male, 27 female). Rats were habituated to the vivarium for a week before the start of any experiments. Animals were pair-housed, unless noted otherwise, in a temperature-controlled vivarium (22 °C) with a 12-h light/dark cycle, lights on at 0700. Food and water were available ad libitum except during Y-maze testing, when rats were food-restricted and weights maintained at a minimum of 90% of free-feeding weights. All experiments were conducted during the light cycle, and with the approval of the Local and University Animal Care Committee at the University of Toronto (AUP 20011310). All experiments were carried out under the guidelines and regulations of the Canadian Council on Animal Care, and reported in accordance with ARRIVE guidelines.

Experimental history

Rats began operant training at approximately 5–6 months of age, following a history of experimentation described previously in detail26. Briefly, they underwent a cued conflict preference assessment in a Y-maze first (Fig. 1B), and experienced prior exposure to ethanol in the homecage via the intermittent access two-bottle choice (IA2BC) paradigm, before the commencement of operant self-administration training.

Fig. 1
figure 1

Experimental timeline and procedures. Timeline of experimental procedures (A). All animals first underwent mixed valence cue conditioning, followed by a cue-conflict preference test in the Y-maze to assess approach-avoidance conflict bias (B). Rats were then pre-exposed to ethanol in their homecages using the Intermittent Access 2 Bottle Choice (IA2BC) paradigm. Next, rats were trained to self-administer ethanol on progressively more effortful schedules of reinforcement and shorter sessions to facilitate bingeing (FR1 2 h, FR1 30 min, FR3 30 min, FR3 15 min). Finally, animals were administered three different tests that capture addiction-like measures (ethanol-seeking with progressive ratio break point (C), extinction (D), and resistance to shock (E)). Animals were given retraining sessions on FR3 15 min schedule between each test.

Full size image

Approach-avoidance conflict Y-maze task

To examine approach-avoidance tendencies under conflict, a mixed-valence conditioning paradigm adapted from Ref.22 was used as described in Ref.26. Rats were habituated, conditioned, and tested in an apparatus that consisted of three arms (50 cm L × 11.5 cm W × 35 cm H) extending from a hexagonal central hub (11.5 cm W × 35 cm H). Guillotine doors separated the central hub from each arm. Arm floors consisted of a shock grid (Lafayette Instrument Co., IN, USA), and contained a well for sucrose delivery at the end.

Rats underwent a total of 4 habituation sessions (5 min). In the first two sessions, rats were first habituated to the whole maze and then to the maze while arms were lined with visuo-tactile denim, tape, and wood bar cues (46 cm L × 9.5 cm H). In the third session, rats were habituated to the conflict test configuration of cues (least and most preferred cues from the second habituation session paired together in one arm, and third set of cues in another arm). Finallly, in the fourth session, animals were habituated to confinement in individual arms with the guillotine doors closing behind them.

Training occurred over eight sessions, and consisted of 2 min forced exposure to each arm, with appetitive cues paired with sucrose delivery (4 × 0.1 mL), aversive cues paired with a mild footshock (4 × 0.28 mA), and neutral cues paired with no outcome. Learning was assessed after the fourth and eighth training session in extinction over 5 min with rats freely exploring all cued arms. After the second acquisition test, rats had one refresher training session before completing the conflict test, also over 5 min in extinction, in which two arms were accessible, one lined with two neutral cues and the other lined with the conflict cues (one sucrose-paired and one footshock-paired). Behavior was live-scored by an experimenter. Time spent in each compartment was measured, and the ratio of conflict arm time to conflict plus neutral arm time was used as the index of conflict preference.

Intermittent access to ethanol two-bottle choice pre-exposure

Rats were given the opportunity to drink 20% v/v ethanol (diluted from 95% v/v in their regular filtered drinking water) freely in their home cages for five weeks, under the intermittent access two-bottle choice (IA2BC) paradigm developed by Ref.47 and modified by Ref.48, as described in Ref.26. Rats had 24-h access to ethanol and water on 3 non-consecutive days each week, with access only to regular drinking water on the other days. This pre-exposure phase to ethanol was included to promote relatively high levels of drinking, and facilitate subsequent operant self-administration48,49.

Operant ethanol self-administration

Apparatus

Testing took place in eight operant chambers (30.5 cm L × 24.1 cm W × 29.2 cm H, Med Associates, VT), each placed within a sound-attenuating box equipped with a ventilation fan. Each chamber had a central magazine on the right side wall that allowed for delivery of 0.1 mL of 20% v/v ethanol via a dipper cup that raised ethanol from a trough positioned below on the outside of the operant box. Two 4 cm wide retractable levers were positioned on the same side wall, 2 cm either side from the magazine and each 5 cm above the floor. The grid floor consisted of parallel bars that connected into a harness that was plugged into a shock generator used to produce mild foot shock. During testing, the ventilation fan was always on and the operant box was illuminated by a 1.8 W 17 V house light on the top central panel on the left side wall of the chamber.

For all operant training sessions, ethanol was available as a 20% v/v solution diluted from 95% v/v in filtered drinking water and placed in a trough placed outside of the operant chamber. The trough was weighed at the beginning (65.0–65.9 g) and end of testing to monitor ethanol consumption during each session. An estimate of ethanol evaporation was measured every four sessions for each operant box by inserting troughs with ethanol underneath the dipper cups and running the program with no animals inside the boxes. This ethanol evaporation was subtracted from all weight change measures to give an accurate recording of how much ethanol the rats consumed in each session. Rats were tested daily at a consistent time of day, and their weights were recorded immediately before each drinking session.

Pre-training

Rats were first exposed to the operant box across three magazine training sessions, in which 0.1 mL of 20% ethanol was delivered by a dipper cup for 15 s on a variable interval 15 s schedule for 30 min. Rats were then introduced to the retractable levers and the instrumental contingency across four 2 h sessions. 0.1 mL of 20% ethanol was delivered on a fixed ratio (FR) 1 schedule. Pressing the active lever resulted in both levers retracting and one delivery of ethanol that lasted 10 s, while pressing the inactive lever led only to the retraction of both levers. Assignment of the side of the active lever (left or right) was counterbalanced across rats.

Binge drinking paradigm

The operant self-administration paradigm used in the current study (Fig. 1A) was based on the binge-drinking model developed on an all-male Long-Evans sample50 and later employed in a study examining sex differences in the microstructure of drinking44. The key features of this paradigm are that the response requirement is increased and then the length of access is shortened across sessions. Following acquisition of the response contingency in the pre-training phase, rats were trained for 7 daily sessions on an FR1 30 min schedule. Ethanol availability was reduced to 5 s at this stage, i.e., the dipper cup remained elevated in the magazine for 5 s, and this duration was held constant through all subsequent training and testing. After the FR1 30 min sessions, the response requirement was increased to FR3 while the session duration remained 30 min. After 10 such sessions, the access time was shortened to 15 min, and rats completed 24 sessions of this final binge-drinking stage on an FR3 15 min schedule.

Progressive ratio

Following self-administration training, rats received a 30 min progressive ratio (PR) test to measure motivation for ethanol (Fig. 1C). The test duration was selected based on the 15–30 min used in a previous binge drinking study50 and was expected to be sufficient to capture the behavior of interest, as responding reportedly plateaus within 10 min50,51,52,53. Ethanol reinforcers were delivered after active lever presses according to the pattern 5e (R*0.2) – 5, where R is equal to the number of reinforcers already earned plus one. The breakpoint was defined as the highest number of presses made to receive a single reward.

Extinction

Following the PR test, rats were retrained on the FR3 15 min schedule for two refresher sessions. Then they were tested on the same schedule for four consecutive sessions, but no ethanol trough was inserted under the dipper cup (Fig. 1D). Therefore, active and inactive lever presses resulted in levers retracting on an FR3 schedule, and active pressing also resulted in an empty dipper cup being raised into the magazine.

Resistance to footshock

Following the final extinction test, rats underwent five refresher FR3 15 min training sessions. Then, to assess compulsive ethanol seeking, a series of five footshock challenge sessions was administered in which a mild footshock (0.20 mA, 0.25 mA, 0.30 mA, 0.35 mA, 0.40 mA, increasing across sessions; Fig. 1E) was delivered at the third second of the dipper cup being raised. Shocks lasted for 0.5 s, and inactive lever pressing resulted in the same retraction of levers on an FR3 schedule.

Blood collection and analysis

Blood was collected from a subset of rats (n = 16, 8 male and 8 female) following regular FR3 15 min training sessions. Briefly, rats were placed in a heated cage following operant training and then anesthetized with isoflurane. Tail vein blood was collected with a 1 mL syringe approximately 30 min after the end of the training session. Blood was immediately frozen at -80 °C and stored in the freezer for later analysis using an alcohol biometer (Albio Meter, PAL, Finland). For analysis, samples were thawed on ice and individual drops of blood pipetted onto the meter with measures collected in duplicates from each sample.

Data analysis

Statistical analyses were performed either in SPSS statistical package version 23.0 (IBM, ON, Canada) or R (version 4.0.3, R Core Team 2020). For all testing, the alpha level was set to 0.05, and two-tailed tests were used. For violations of sphericity, a Huyhn-Feldt correction was used. Bonferroni corrections were used as appropriate for multiple comparisons.

Active and inactive lever presses were recorded from all operant sessions. Ethanol consumption, where appropriate, was calculated as grams of pure ethanol per kilogram of body weight, using densities of 0.789 g/ml and 1 g/ml for ethanol and water, respectively. Percent consumption was calculated based on the volume of ethanol consumed compared to the volume of ethanol delivered. For comparisons across operant schedules, averaged session values across the final four days of testing at each schedule were used. Repeated measures (RM) ANOVA were used with sex as a between subject and schedule as a within subject variable to compare lever pressing and percent consumption at each schedule of reinforcement.

To assess drinking patterns at each operant schedule, lever pressing data (percent of total presses) and consumption data (g/kg) were analyzed in one-minute bins. Consumption by bin was approximated by assuming that drinking was distributed equally across all ethanol deliveries. Importantly, only deliveries that were accompanied by a head entry into the magazine were included in this approximation. Separate RM ANOVA were run at each operant schedule with sex as a between subject and bin as a within subject variable. Ethanol taking concentrated at the onset of an access session is not thought to be meaningful when overall consumption is low43, thus we derived from these measures two indices of frontloading: onset skew, the cumulative percentage of presses completed after five minutes of ethanol access, and bingeing, the cumulative ethanol consumption within five minutes of ethanol access.

Consumption was subsequently plotted cumulatively to visualize against an estimated metabolic rate for ethanol (0.007 g/kg/hr, see Refs.54,55,56). Blood ethanol concentrations were assessed during the final FR3 15 min schedule only, and values obtained from the meter were correlated with consumption during the session immediately prior to blood collection.

To assess sex differences in conflict behavior and addiction like behavior, t-tests (cued-conflict preference and breakpoint from progressive ratio) or RM-ANOVA (lever pressing and outcomes across 30 min of PR, extinction across four sessions, resistance to footshock across five sessions) were used. Extinction pressing and ethanol consumption in the face of footshock were both represented as a percentage of baseline (from the FR3 15 min) for each individual subject. Values across the extinction and footshock sessions were subsequently summed to generate singular extinction and footshock scores, which were then compared by sex in t-tests.

Multiple regression analyses were carried out to regress addiction criteria measures (breakpoint, extinction score, and footshock score) onto all the remaining drinking measures – FR3 15 min consumption, extinction score, footshock score, breakpoint, onset skew, bingeing, onset skew by bingeing interaction – plus sex (coded as female = 0, male = 1), conflict preference, and a sex by conflict preference interaction term (Table 1). All individual, non-interaction terms that were included in the models are shown in a correlation matrix (Table S1). All predictors were group-centered by sex. Final multiple regression models were reached by removal of the most insignificant predictors (by p-value) one by one until all remaining predictors had p < 0.05 or were included in significant interactions, or until the adjusted r square value decreased. Predictors were not eligible for removal if their interaction terms remained in the model. If removing an insignificant predictor lowered the adjusted r square measure, then that predictor was returned to the model and that version deemed final. To assess whether using a multiple regression approach was appropriate, we first visually verified a linear relationship between predictors and dependent variables for each regression analysis. Upon reaching each final model, we verified on a histogram that residuals were normally distributed and confirmed homoscedasticity by plotting the residuals against predicted values. Finally, we used Variance-inflation-factor (VIF) values to check for multicollinearity (with values above 10 deemed intolerable, and above 5 deemed moderate but tolerable) and the Durbin-Watson statistic to test for autocorrelation of residuals (with values between 1.5 and 2.5 deemed acceptable).

Table 1 Multiple regression predictors.
Full size table

Simple slopes analyses57,58 were used to follow up on significant interactions in multiple regression models. In the case of continuous predictors, slopes were investigated at three levels of interest: the mean, plus one standard deviation above and below. In the case of the binary sex predictor, slopes of female and male values were investigated.

Results

The timeline of our experimental procedures is outlined in Fig. 1. Rats first completed a Pavlovian approach-avoidance conflict Y-maze paradigm in which their approach-avoidance bias for conflicting cues (predicting shock and sucrose) versus neutral cues was assessed. Subsequently, rats were trained on an operant binge drinking paradigm as previously reported44,50. Patterns of voluntary self-administration were analyzed across operant schedules that increased the response requirement for ethanol and then compressed access time50. To assess a range of addiction-like drinking behaviors, rats were finally tested on a progressive ratio challenge, four consecutive days of extinction testing, and a series of footshock challenges (0.2–0.4 mA).

Body weights increased across time faster in males than females

To provide proper context for weight-dependent measures of intake, an analysis of body weights was conducted as a RM-ANOVA across schedule (FR1 30 min, FR3 30 min, FR3 15 min, Fig. 2A) with sex as a between-subjects factor and body weight averaged across the final 4 days of drinking at each schedule, to match the days of analysis of drinking. Males outweighed females as expected (Sex: F(1,51) = 534.236, p < 0.001, ηp2 = 0.913), and weights increased across time (Schedule: F(1.254,63.935) = 406.942, p < 0.001, ηp2 = 0.889) in a sex-dependent manner (Sex x Schedule: F(1.254,63.935) = 76.888, p < 0.001, ηp2 = 0.601). Follow up analyses revealed that males outweighed females at each schedule (all p < 0.001) and that each sex gained weight from one schedule to the next (all p < 0.001). Furthermore, higher rate of weight gain occurred in males compared to females, (females: 6.9% change (293.991 g to 314.324 g), males: 11.3% change (461.260 g to 513.365 g) from FR1 30 min to FR3 15 min schedule).

Fig. 2
figure 2

Ethanol taking and seeking measures. Body weights (A), ethanol consumption (g/kg; (B), ethanol consumption (percent of delivered ethanol consumed; (C), number of delivered outcomes (D), number of deliveries during which no head entry was made (E), latency to head entry after ethanol delivery (F), active (G) and inactive (H) lever presses, and inactive lever presses (percent of total lever presses; (E) across the three operant schedules in males and females. Data plotted are mean values from the final four sessions at each schedule ± SEM. ***indicates significant main effect of sex at p < .001. Data points labeled with different letters (a,b,c) denote statistically significant differences between schedules as determined by post-hoc tests following significant main effects of schedule. Data points sharing the same letter are not significantly different. 

Full size image

Ethanol consumption increased across schedules as response requirement was increased and access time was reduced

To determine the influence of changing operant testing parameters on ethanol-taking, RM-ANOVA was used to assess the effects of sex (between subject) and schedule (within subject) on ethanol consumption, number of ethanol outcomes delivered, percent of deliveries consumed, number of deliveries not accompanied by a head entry, latency to nosepoke, active and inactive lever presses, and percent of inactive lever presses (Fig. 2B-I). Ethanol consumption, as measured in grams of pure ethanol per kilogram of body weight, increased over time (Schedule: F(1.840, 93.842) = 8.639, p < 0.001, ηp2 = 0.145; Fig. 2B), with the FR1 30 min schedule yielding the lowest consumption levels when compared to the others: p < 0.01 against FR3 15 min and p < 0.05 against FR3 30 min, p = 0.183 between the two FR3 schedules. As expected, females drank more than males across all schedules (Sex: F(1, 51) = 48.565, p < 0.001, ηp2 = 0.488) and escalated their drinking significantly more across schedules than in males (Sex x Schedule: (F(1.840, 93.842) = 3.496, p < 0.05, ηp2 = 0.064). Females not only outdrank males at each schedule (all p < 0.001) but also exhibited their lowest consumption in the FR1 30 min schedule (p < 0.001 against both FR3 schedules, with the two FR3 schedules no different p = 0.47). In contrast, male consumption never changed with schedule (all p > 0.62).

Interestingly, the increase in ethanol consumption with each schedule was found to be due to all rats consuming a greater percentage of delivered ethanol in each successive schedule (F(1.885, 96.114) = 30.451, p < 0.001, ηp2 = 0.374; Sex: F(1,51) = 0.928, p = 0.340, ηp2 = 0.018; Sex x Schedule: F(1.885, 96.114) = 0.723, p = 0.480, ηp2 = 0.014; Fig. 2C), with the FR3 15 min schedule yielding the highest percent consumption, followed by FR3 30 min and FR1 30 min the lowest (all p < 0.001), rather than rats pressing for more ethanol deliveries at each successive schedule, as the number of outcomes earned did not differ by schedule (F(2,102) = 0.713, p = 0.493, ηp2 = 0.014; Fig. 2D), nor by sex (F(1,51) = 2.994, p = 0.090, ηp2 = 0.055), with no interaction (F(2,102) = 0.449, p = 0.640, ηp2 = 0.009).

Consistent with the pattern of stable deliveries but increasing percent consumption, rats failed to make magazine entry during ethanol delivery more at the earlier schedules: (F(1.426,72.704) = 37.415, p < 0.001, ηp2 = 0.423; Fig. 2E) as the increase in response requirement drove a reduction in head-entry omissions (FR3 15 min and FR3 30 min both p < 0.001 against FR1 30 min, but no different from each other, p = 0.219). There was no effect of sex (F(1,51) = 2.817, p < 0.099, ηp2 = 0.052) or interaction between sex and schedule (F(1.426,72.704) = 2.905, p < 0.078, ηp2 = 0.054), though the trend was for females to omit more head entries during delivery at FR1 30 min, with this pattern reversing by FR3 15 min with males omitting more at that schedule. As ethanol deliveries were more reliably followed by head entries across schedules, so too did head entries occur faster, with an effect of schedule on head-entry latency (F(1.684,85.860) = 19.112, p < 0.001, ηp2 = 0.273; Fig. 2F). This was driven by access compression, with post-hoc testing revealing the FR3 15 min schedule to have the fastest head entries (p < 0.001 against both 30 min schedules, which did not differ from each other, p = 0.214). There was a trending effect of sex, with females faster than males (F(1,51) = 3.640, p < 0.062, ηp2 = 0.067) but no interaction between sex and schedule (F(1.426,72.704) = 0.333, p < 0.680, ηp2 = 0.006.)

Although the number of delivered outcomes did not change across schedules, the increasing response requirement from FR1 to FR3 yielded a main effect of schedule on raw lever presses (Active: F(2,102) = 155.221, p < 0.001, ηp2 = 0.753; Inactive F(1.668, 85.050) = 5.116, p < 0.05, ηp2 = 0.091; Fig. 2G-H). Active lever pressing was higher on both FR3 schedules compared to FR1 (both p < 0.001) but did not differ between FR3 30 min and FR3 15 min (p = 0.971). Inactive lever pressing, however, was increased from FR1 30 min only at the FR3 30 min schedule (p < 0.05) while the FR3 15 min inactive lever pressing was no different from FR1 30 min (p = 0.087) or FR3 30 min (p = 0.426). There was no influence of sex (Active: F(1,51) = 3.288, p = 0.076, ηp2 = 0.061; Inactive: F(1,51) = 2.704, p = 0.106, ηp2 = 0.050) or an interaction with schedule (Active: F(2,102) = 1.433, p = 0.243, ηp2 = 0.027; Inactive: F(91.668,85.050) = 0.424, p = 0.619, ηp2 = 0.008). To assess the relative distribution of effort between the two levers, we calculated the percentage of total lever presses made on the inactive lever, and we found that this decreased across schedules (F(2, 102) = 19.483, p < 0.001, ηp2 = 0.276; Fig. 2I), with both FR3 schedules yielding a reduced percentage of pressing on the inactive lever compared to FR1 30 (both p < 0.05). The two FR3 30 min schedules were no different (p = 0.241). There was no effect of sex (F(1, 51) = 0.050, p = 0.825, ηp2 = 0.001) or interaction (F(2, 102) = 0.234, p = 0.792, ηp2 = 0.005).

Sex differences in frontloading: Females binged more than males, with no difference in onset skew at FR1 30 min

To assess the pattern of ethanol seeking and taking at each operant schedule, we ran separate ANOVAs to compare the effects of sex (between subject) and bin (1 min duration, within subject) on percent pressing and consumption under FR1 30 min, FR3 30 min and FR3 15 min. Subsequently, onset skew (cumulative percent pressing achieved through 5 min) and bingeing (cumulative consumption through 5 min) were compared by sex to provide two different indices of frontloading at each schedule.

At the initial FR1 30 min schedule, males and females displayed a similar decline in pressing activity over time (Fig. 3A), as evidenced by a main effect of bin (F(4.525,235.563) = 106.220, p < 0.001, ηp2 = 0.676) but no interaction with sex (F(4.525,235.563) = 1.455, p = 0.211, ηp2 = 0.028). Consumption likewise followed a declining pattern across bins (F(3.878,197.779) = 97.401, p < 0.001, ηp2 = 0.656). Rats completed about 62% of their total active lever presses on the FR1 30 min schedule within the first five minutes (Fig. 3A inset), a disproportionate amount of pressing compared to elapsed time (against 16.67, t(52) = 23.829, p < 0.001), and therefore indicative of frontloading as a group. There was no sex difference in onset skew (t(51) = -0.263, p = 0.794).

Fig. 3
figure 3

Frontloading patterns. Active lever presses per bin (as a percent of total active lever presses) and ethanol consumption per bin (g/kg) in males and females in the FR1 30 min (A, B), FR3 30 min (C, D), and FR3 15 min (E, F) schedules. Inset: cumulative percent pressing completed in the first five bins (a, c, e) and cumulative ethanol consumption (g/kg) achieved in the first five bins (B, E, F). * p < .05, ** p < .01, *** p < .001. & indicates a male score significantly greater than female score and # a female score significantly greater than a male score as determined by post-hoc testing to follow up on a sex by bin interaction. All data expressed are mean ± SEM taken from the final four days of testing at each schedule.

Full size image

Females, however, outdrank males (F(1,51) = 21.997, p < 0.001, ηp2 = 0.301; Fig. 3B) and reduced their consumption more sharply from a higher peak compared to males (Sex x Bin: F(3.878,197.779) = 7.971, p < 0.001, ηp2 = 0.135). Post hoc testing showed higher consumption in females compared to males at bins 1–4, 8, 11, 23, and 27–29 (all p < 0.05). This pattern held in the first five minutes of the session, when females binged more than males (t(51) = 4.261, p < 0.001; Fig. 3B inset).

Females binged more than males, while male pressing was skewed towards onset more than female pressing at FR3 30 min

With the response requirement increasing to FR3 and access time remaining at 30 min, males and females continued to be more active early on in sessions, with percent pressing decreasing across bins (F(4.085,290.440) = 158.814, p < 0.001, ηp2 = 0.757), but no sex by bin interaction (F(4.085,290.440) = 1.945, p = 0.103, ηp2 = 0.037; Fig. 3C). As a group, rats completed about 65% of their total active lever presses on the FR3 30 min schedule within the first five minutes, indicative of frontloading and disproportionately greater than elapsed time (against 16.67, t(52) = 25.450, p < 0.001). Males had a higher onset skew, indicating more concentrated pressing in the first sixth of the access session, (t(51) = -2.150, p < 0.05; Fig. 3C inset).

Similar to percent pressing, consumption declined across bins (F(5.787,295.159) = 134.108, p < 0.001, ηp2 = 0.724), and as in the FR1 30 min schedule females outdrank males overall (F(1,51) = 44.955, p < 0.001, ηp2 = 0.469; Fig. 3D). A sex by bin interaction (F(5.787,295.159) = 8.557, p < 0.001, ηp2 = 0.144) revealed a steeper decline in female consumption from a higher starting point relative to males. Post-hoc testing showed higher consumption in females compared to males at bins 1–9, 11–14, 17–19, and 22–24. The calculated bingeing for animals from the first five minutes of the session showed that females reached higher levels of consumption in that time (t(51) = 6.069, p < 0.001; Fig. 3D inset).

Females binged more than males, but male pressing was skewed towards onset more than female pressing at FR3 15 min

The final operant schedule maintained the FR3 response requirement while access time was reduced to 15 min to induce bingeing50. At this schedule, as in the others, drinking activity was most concentrated early in the session and a main effect of bin showed decreased percent pressing across bins (F(3.986,207.688) = 230.835, p < 0.001, ηp2 = 0.819; Fig. 3E). Here a sex by bin interaction revealed a steeper decline in males from a higher starting point early in the sessions (F(3.986,207.688) = 7.027, p < 0.001, ηp2 = 0.121): males concentrated a higher percentage of their pressing in the first two bins compared to females, while females subsequently maintained higher relative pressing through the middle bins 5–8 of the session (all p < 0.05). The calculated onset skew revealed a higher overall relative concentration of pressing in males within the first third of the access session (t(51) = -3.143, p < 0.01; Fig. 3E inset), with all rats together completing a disproportionate 75% of their pressing in the first third of the session (against 33.33, t(52) = 57.662, p < 0.001).

As with pressing, consumption declined across bins on the FR3 15 min schedule (F(4.118,210.004) = 179.002, p < 0.001, ηp2 = 0.778; Fig. 3F). Females outdrank males (F(1,51) = 39.783, p < 0.001, ηp2 = 0.438), with the same sex by bin interaction as on the earlier schedules (F(4.118,210.004) = 8.817, p < 0.001, ηp2 = 0.147) revealing that female consumption started higher and reduced more sharply compared to male consumption. Post-hoc testing showed higher consumption in females compared to males at bins 1–9, 11, and 14. Within the first 5 min only, calculated bingeing values were higher for females compared to males (t(51) = 5.495 p < 0.001; Fig. 3F inset).

Blood ethanol concentrations correlated well with ethanol intake

To assess the pharmacological relevance of the amounts of ethanol consumed in the study, we first plotted consumption as a cumulative curve against an estimated metabolic rate (0.007 g/kg/hr54,55,56; Fig. 4A-C). From the culminating FR3 15 min sessions, blood was collected from a subset of animals (n = 8 each male and female) for BEC analysis. Drinking in these sessions correlated well with BEC at r2 = 0.800, p < 0.001 (Fig. 4D). Females, exceeding the binge threshold of 80 mg/dL on average, reached higher BECs than males, who fell short of that threshold (Females, mean: 1.194 mg/dL, SD: 0.375; Males, mean: 0.602 mg/dL, SD: 0.354, t(14) = 3.247, p < 0.01; Fig. 4E). Sex differences in BECs were consistent with higher consumption in females (mean: 1.075 g/kg, SD: 0.444) compared to males (mean: 0.693 g/kg, t(14) = 3.809, p < 0.01; Fig. 4F). Consumption measured on blood collection day did not differ from overall consumption on the FR3 15 min schedule reported in the study for females (t(33) = 1.692, p = 0.100) or males (t(32) = 0.649, p = 0.521.

Fig. 4
figure 4

Ethanol consumption and blood ethanol concentrations. Ethanol consumption (g/kg) on the FR1 30 min (A), FR3 30 min (B) and FR3 15 min (C) schedules plotted as cumulative consumption against an estimated metabolic rate of 0.007 g/kg/min. All data expressed are mean ± SEM taken from the final four days of testing at each schedule. Blood ethanol concentrations plotted against consumption on the day of blood collection (D) and comparing males against females (E). Ethanol consumption comparing males against females on the day of blood collection (F). ** p < .01.

Full size image

Females were more conflict-avoidant than males

From the conflict test on the Y-maze, calculated cued-conflict preference scores were higher for males compared to females (t(51) = -2.622, p < 0.05; Fig. 5A).

Fig. 5
figure 5

Addiction-like measures. Cued-conflict preference as assessed on the Y-maze task prior to operant ethanol self-administration (A). Highest response requirement completed by rats on the progressive ratio task (B). Active lever presses (C) and number of ethanol reinforcers delivered (D) by minute during the progressive ratio task. Active lever presses across 4 extinction sessions, expressed as a percent of average lever presses during training (E). Ethanol consumption (g/kg) across 5 levels of footshock expressed as a percent of average consumption during training (F). Inset, summed scores across extinction (E) and footshock (F) sessions). * p < .05, *** p < .001. # indicates a female score significantly greater than a male score as determined by post-hoc testing to follow up on a sex by bin interaction. All data expressed are mean ± SEM.

Full size image

Females displayed a more addiction-like profile than males

Progressive ratio test of motivation

On the progressive ratio test of motivation, females reached higher breakpoints than males (t(51) = 2.444, p > 0.05; Fig. 5B). To verify that 30 min was sufficient for rats to have reached their breakpoints, we plotted the decline in lever pressing (F(7.358,367.901) = 53.729, p < 0.001, ηp2 = 0.518; Fig. 5C) and outcomes (F(2.887,144.346) = 162.861, p < 0.001, ηp2 = 0.765; Fig. 5D) across time. On both measures, post-hoc testing revealed that the fourth minute was the latest to differ from any subsequent minute (all p > 0.067), revealing no changes in lever pressing or outcomes delivered across the final 26 min.

Ethanol seeking despite its unavailability

Pressing on the active lever during extinction was expressed as a percentage of baseline FR3 15 min pressing, and, indicative of extinction, it declined across the four sessions (F(2.380,121.378) = 56.482, p < 0.001, ηp2 = 0.526; Fig. 5E). Pairwise comparisons revealed significant drops in pressing from sessions 1 to 2 and again from 3 to 4 (both p < 0.05), while pressing was comparable between the middle two sessions (p = 0.427). There was no main effect of sex (F(1,51) = 2.004, p = 0.163, ηp2 = 0.038) though a sex by session interaction (F(2.380,121.378) = 3.483, p < 0.05, ηp2 = 0.064) revealed somewhat more addiction-like behavior in more persistent pressing in females, who pressed at a higher percent of baseline than males on the first and the last sessions (both p < 0.05). Overall, when extinction pressing values were summed into a singular score, there was no sex difference (t(51) = 1.416, p = 0.163; Fig. 5E inset).

Ethanol taking in the face of mild footshock punishment

When rats were tested on the FR3 15 min schedule with a mild footshock delivered on the third second of each ethanol delivery, footshock intensity was increased from 0.20 to 0.40 mA across 5 sessions. Ethanol consumption, as a percentage of baseline FR3 15 min consumption, declined across sessions, showing a reduction in compulsive ethanol-taking as the intensity of punishment increased (F(3.555,181.322) = 54.349, p < 0.001, ηp2 = 0.516; Fig. 5F). Pairwise comparison showed successive session over session declines (all p < 0.05) with the exception of session 3 to 4 (0.30 to 0.35 mA intensity footshocks; p = 0.127). There was no main effect of sex on consumption (F(1,51) = 0.812, p = 0.372, ηp2 = 0.016), but a sex by session interaction revealed distinct patterns of declining consumption in males and females (F(3.555,181.322) = 3.031, p = 0.023, ηp2 = 0.056). Post hoc testing showed that males and females never differed in percent consumption at any shock intensity (all p > 0.107). Instead, we observed that male pressing declined more rapidly than female pressing in the face of increasing footshock (male session 1 > 2 and 2 > 3, both p < 0.001; female session 1 = 2 and 2 = 3, both p > 0.06), suggesting a somewhat more footshock-resistant and compulsive profile in females. Overall, when footshock consumption values were summed into a singular score, there was no sex difference (t(51) = 0.901, p = 0.372; Fig. 5F inset).

Frontloading and cued-conflict preferences can be used in predicting addiction-like drinking measures

Multiple regression models were used to examine the effects of sex, approach-avoidance conflict bias, bingeing and drinking measures on addiction-like drinking outcomes using three challenge measures: progressive ratio, extinction, and resistance to footshock (Table 1). The dependent measure for motivation was the breakpoint reached on the progressive ratio task, for extinction it was the singular Ext_Score derived from the sum of pressing (percent of baseline) across all four sessions, and for resistance to footshock it was the FS_score derived from the sum of consumption (percent of baseline) across all five sessions. All independent (i.e. non-interaction) terms used in the models are shown in a correlation matrix by sex (Table S1). All final models were checked for multicollinearity and autocorrelation in the residuals and deemed satisfactory (see Table S2 for VIF values and Durbin-Watson statistics).

Motivation to seek ethanol is predicted by sex, frontloading measures, and persistent ethanol-seeking in extinction

The final model to predict breakpoint was as follows:

BP = 29.090—6.849*Sex—0.332*Onset_Skew + 16.023*Binge – 0.920*Onset_Skew*Binge + 0.015*Ext_Score.

This model significantly predicted breakpoint (F(5,47) = 7.469, p < . 001), with an r2 of 0.443 (adjusted r2 = 0.383), which indicates that 44% of the variance in breakpoint was explained by the predictors (Fig. 6A).

Fig. 6
figure 6

Multiple regression analyses. Plots of predicted dependent variables (DVs) from the multiple regression models against observed DV data for breakpoint (A), extinction score (C) and footshock score (E). Simple slopes were visualized to resolve interactions in the final multiple regression models. Frontloading and bingeing were considered at average values and 1 standard deviation above and below in predicting breakpoint (B). Sex was considered at male and female across average cued-conflict preference scores and 1 standard deviation above and below in predicting extinction (D) and footshock (F). Bands represent 95% confidence intervals around predicted scores.

Full size image

Sex, with higher predicted breakpoints for females (t(47) = -3.431, p < 0.01),

Onset_Skew (t(47) = -3.516, p < 0.001), Binge (t(47) = 3.706, p < 0.001), and the Onset_Skew*Binge interaction term t(47) = -2.117, p < 0.05) were found to be significant predictors of breakpoint.

Simple slopes analyses revealed the slopes of Onset_Skew (Fig. 6B) were significantly negative at both high (-0.597, t(47) = -3.459, p < 0.01) and average Binge (-0.332, t(47) = -3.516, p < 0.001), but there was no effect of Onset_Skew at low Binge (-0.067, t(47) = -0.481, p = 0.633).

Extinction of ethanol seeking is predicted by conflict preference in a sex-dependent manner, as well as ethanol consumption levels during FR3 15 min sessions, frontloading measures, and resistance to footshock

The final model to predict extinction of ethanol seeking was as follows:

Ext_Score = 249.679 – 30.416*Sex – 229.245*FR3_15_min_con + 0.329*FS_score – 3.486*Onset_Skew + 186.088*Binge + 107.636*Conflict_Pref – 432.402*Sex*Conflict_Pref.

This model significantly predicted Ext_score (F(7,45) = 6.172, p < 0.001) with an r2 of 0.490 (adjusted r2 = 0.410; Fig. 6C).

FR3_15_min_con (t(45) = -3.399, p < 0.01), FS_Score (t(45) = 4.836, p < 0.001), Onset_Skew (t(45) = -3.255, p < 0.01), Binge (t(45) = 2.337, p < 0.05) and Sex*Conflict_Pref (t(45) = -3.071, p < 0.01) were all significant predictors of Ext_Score, while the Sex term (t(45) = -1.826, p = 0.074) was marginally significant, suggesting less persistent pressing in extinction in males, and the Conflict_Pref term was insignificant (t(45) = 1.349, p = 0.184) but remained in the model as it was part of the interaction.

To follow up on the interaction, simple slopes were analyzed to show the influence of conflict preference separately on females and males (Fig. 6D). Conflict preference did not influence predicted Extinction Scores in females (B = 107.636, t(45) = 1.349, p = 0.184), but it negatively predicted Extinction Scores in males (B = -23.766, t(45) = -2.776, p < 0.01).

Ethanol taking in the face of punishment is predicted by conflict preference in a sex-dependent manner, as well as onset skew and ethanol seeking during extinction

The final model to predict ethanol seeking and taking in the face of punishment was as follows:

FS_Score = 187.384 – 18.094*Sex + 279.957*FR3_15_min_con + 0.973*Ext_Score + 5.637*Onset_Skew – 158.825*Binge -24.042*Conflict_Pref + 610.914*Sex*Conflict_Pref + 11.795*Onset_Skew*Binge.

This model significantly predicted FS_Score (F(8,44) = 4.733, p < 0.001) with an r2 of 0.463 (adjusted r2 = 0.365; Fig. 6E).

FR3_15_min_con (t(44) = 2.200, p < 0.05), Ext_Score (t(44) = 4.562, p < 0.001), Onset_Skew (t(44) = 2.797, p < 0.01) and Sex*Conflict_Pref (t(44) = 2.424, p < 0.05) all significantly influenced predicted FS_Score. Sex (t(44) = -0.602, p = 0.550), Binge (t(44) = -1.094, p = 0.280, conflict preference (t(44) = -1.70, p = 0.866) and Onset_Skew*Binge (t(44) = 1.767, p = 0.084) remained in the model despite not being significant.

Simple slopes were calculated to follow up on the Sex by Conflict Preference interaction, revealing no influence of conflict preference in females (B = -24.042, t(44) = -1.70, p = 0.866) but a significantly positive predictive influence of conflict preference on FS_Score in males (B = 586.872, t(44) = 2.925, p < 0.01; Fig. 6F).

Discussion

In the current study, we aimed to assess the influence of drinking patterns and cued-conflict preference in predicting addiction-like drinking outcomes following training on a binge-drinking paradigm. We found ethanol consumption to escalate across the three schedules of operant training despite growing body weights. We detected the emergence of sex differences in drinking patterns, with females outdrinking males reliably but males developing more concentrated ethanol seeking at the onset of access sessions (high onset skew). Following the FR3 15 min training, females displayed a more addiction-like profile than males, reaching higher breakpoints than males and exhibiting somewhat greater persistence in pressing for ethanol in extinction and resistance to reducing ethanol consumption when deliveries were accompanied by footshock. Finally, multiple regression models implicated frontloading and cued-conflict preference to predict addiction-like drinking outcomes in unexpected ways. Bingeing predicted higher breakpoints, but this effect was selectively present at average and lower onset skews, with higher onset skews yielding lower breakpoints. Cued-conflict preference interacted with sex, negatively predicting extinction and positively predicting footshock only in males. Together, our results highlight drinking patterns as complex predictors of addiction-like drinking outcomes, and implicate approach-avoidance conflict behavior as a sex-dependent predictor of addiction-like drinking outcomes.

Operant behavior becomes more ethanol-directed along the progression to a schedule with shorter access

Ethanol consumption was found to increase across schedules despite growing body weights. We analyzed weight-independent measures of ethanol-seeking to investigate what behavior might underlie the escalating intake. Our results indicated that in addition to gaining weight, rats failed to lever press for more ethanol deliveries across schedules, intensifying the conundrum of how consumption levels might be escalating. What we found was an increase in consumption of the delivered ethanol, from just over 70% at the initial FR1 30 min schedule to almost 90% at the FR3 15 min schedule. Such a measure is rarely, if ever, reported in the literature, as sometimes details on how consumption is assessed are not included (e.g. Refs50,59) and, in fact, all available ethanol is assumed to be consumed in other reports (e.g. Ref60). In our approximation of minute-by-minute ethanol consumption, our overall consumption estimates were distributed equally across ethanol deliveries, and excluded those during which rats never made a head entry. We observed that these head entry omissions were most common at the initial FR1 30 min schedule, helping to explain the lower percentage of consumption. In contrast, head entries into the magazine followed the ethanol delivery most closely on the final FR3 15 min schedule, demonstrating a heightened eagerness to consume the reward. Consistent with these observations, and indicative of lever pressing becoming a more ethanol-directed behaviour across schedules: while active lever pressing increased from FR1 30 min to FR3 30 min and remained elevated at FR3 15 min, presses on the inactive lever saw a similar jump from FR1 30 min to FR3 30 min, but then returned to levels no different from FR1 30 min on the final FR3 15 min schedule. Similarly, we report a decrease in the percentage of lever presses being made on the inactive lever: this percentage dropped from nearly 20% on the FR1 30 min schedule to below 10% on the FR3 15 min schedule. In summary, ethanol deliveries were sought more reliably and more quickly across schedules, and a reduced proportion of lever presses were made on the lever that yielded no outcome. This resulted in an increasing consumption rate of the ethanol outcome that appears to stem from lever pressing that shifted in male and female rats to become more ethanol-directed. To our knowledge, this is the first demonstration of such a shifting pattern, although our findings may be related to increased motivational breakpoints detected by Ref50 across these operant schedules.

One possibility is that the length of exposure to ethanol over time, and more specifically, metabolic tolerance, may have increased the desired intake of ethanol levels. Rats have been shown to modulate their intake to reach consistent BECs61, suggestive of drinking for the pharmacokinetic effect, which would change with tolerance. However, while chronic tolerance could develop across the timeframe of our study, it typically occurs with higher levels of ethanol exposure and is not thought to specifically influence frontloading43,62. Additionally, prior research has shown increased motivational breakpoint on the FR3 15 min schedule compared to schedules with longer access times when ethanol history is better controlled, demonstrating increased motivational drive for ethanol50. Thus, the development of chronic tolerance as a between-schedule confound in the current study is not the most parsimonious explanation; rather, the increased response requirement from FR1 to FR3 adds to the cost of lever pressing and may facilitate this change, which could then be further facilitated by access compression, as undirected lever pressing becomes more costly in terms of time. Therefore, we expect this shift to more ethanol-directed behavior is at least facilitated by the FR3 15 min binge-drinking schedule.

The proposed binge-drinking paradigm yields high levels of ethanol intake, with females outdrinking males

Females consumed more ethanol than males across all operant schedules, consistent with what is widely reported26,40,68,69,70 but see Refs.44,71. Escalation of ethanol intake was observed across operant schedules, similar to the findings between different schedules in a previous study proposing this binge-drinking paradigm50. Drinking at each schedule easily outpaced an estimate of metabolic rate, suggesting consumption of pharmacologically relevant levels of ethanol. BECs correlated well with consumption, and our rats on average reached binge-level consumption (80 mg/dL)43,50 per their blood ethanol concentrations, although low-drinking individuals, and males on average, fell short of this mark. It is of note that we measured lower BECs in males compared to previous reports of male BECs (Long Evans) on this paradigm50. The females reached higher BECs than males, falling within a range that others have previously reported on voluntary self-administration paradigms in Long-Evans female rats72,73. Overall, the ranges of drinking levels and BECs observed suggest this training promotes high drinking and facilitates detection of individual differences.

Sex differences in frontloading depend on the measure assessed, but do not explain differences in ethanol consumption

Frontloading is a commonly reported phenomenon, but gaps in the literature remain due to a wide range of parameters and measures used to assess frontloading, as well as limited work evaluating sex differences (reviewed in Ref.43). Indeed, many frontloading studies rely on males and utilize paradigms with long access paradigms, such as drinking in the dark (DID, 2–4 h; e.g. Refs.63,64) or IA2BC (24 h; e.g. Refs.65,66). These investigations have also been designed to detect the presence of frontloading in one group (by comparison to a water group or by early session intake compared to later) rather than to explore subtle differences between multiple ethanol drinking groups60,64,65,66. We assessed two features of frontloading separately: onset skew, a measure of how concentrated ethanol-seeking is early on in an access session, and bingeing, a raw, weight-dependent measure of ethanol intake early on in an access session. As expected, we found that females binged at higher levels than males (see mixed findings in Ref.44), and similar to others, we observed a disproportionate level of ethanol seeking in the first 5 min of access sessions43,44,50. However, an unexpected sex difference emerged: once the response requirement was increased to FR3, males began to concentrate a greater percentage of their ethanol seeking in the first 5 min of access sessions compared to females. Unlike the measure of bingeing, which suggested females frontload more, this measure of onset skew suggested males frontload more. It has been proposed that females frontload more, and that this is the cause of their overall higher consumption43,44, but our data do not support this interpretation.

In the present study, females outdrank males, regardless of differences in drinking patterns and despite equivalent levels of ethanol seeking. The reasons for this are beyond the scope of the study, but likely involve reported sex-specific differences in ethanol pharmacokinetics (e.g., metabolism, hormone influences)74,75,76,77 and potential differences in neural circuit mechanisms78,79.

The FR3 15 min schedule yields a more addiction-like phenotype in females than males

Preclinical research has generally demonstrated sex differences in which females are more vulnerable to addiction than males (reviewed in Refs.30,31,80). Our data add support to the interpretation the females are more vulnerable to addiction-like phenotypes, and lend validity to the proposed binge-drinking paradigm in facilitating addiction-like behavior. Our females reached higher breakpoints than males, although previous research has only sometimes produced this finding60 and often suggested breakpoints do not differ by sex59,81,82,83 or are even higher in males67. Pressing plateaued within 5 min on our PR test, consistent with other reports suggesting that breakpoints can be detected within short periods50,51,52,53. While we found no main effect of sex on extinction pressing or ethanol consumption in the face of footshock, we found more persistent pressing across extinction sessions (similar to Ref.68 but see Ref.83,84 for no sex difference) and slowed reduction in consumption across increasing footshock intensities in females compared to males, consistent with aversion-resistant drinking paradigms using footshock punishment39,67.

Combination of two indices of frontloading predicts addiction-like phenotypes

In humans, as with other drinking measures, rates of bingeing are rising rapidly in females than males, yielding more negative consequences in females7,46. Thus, the assessment of frontloading preclinically may provide insights into drinking patterns promoted by concepts such as happy hour44,45 where women tend to be more susceptible to increasing their drinking, and more vulnerable to negative consequences45. In our study, we used two indices of frontloading, bingeing and onset skew, in multiple regression analyses to predict addiction-like drinking behavior. In specifying these two measures, we hoped to be able to detect possibilities beyond the suggested simple relationship that high frontloading (or bingeing) reflects high motivation for alcohol43. Consider four drinking scenarios over four hours, beginning with an initial happy hour. In the first scenario, one drink occurs only within the first hour, showing high onset skew without binge-level intake. In the second, an initial drink occurs in the first hour and another later, with lower onset skew and no bingeing. In the third scenario, 4 drinks are concentrated in the first hour only, reflecting high onset skew and bingeing. In the fourth pattern, 4 drinks occur in the first hour and another 4 drinks later, indicating an equal level of bingeing but low onset skew due to more evenly distributed consumption. This last pattern of drinking represents the riskiest form of drinking, and yet this pattern combines high bingeing with low onset skew – a phenotype combination that may be overlooked in unidimensional assessments of frontloading. Indeed, relying on indices of frontloading based solely on intake over a fixed time suggests that higher frontloading predicts higher motivation, but, using measures proportional to whole access sessions may suggest the opposite. Our results reveal that a high onset skew represents less persistent ethanol-seeking, and that an earlier termination of ethanol-seeking on training sessions is linked with a lower motivational breakpoint. In contrast, the rats most prone to high breakpoints are the high binge, low onset skew rats who consume at high levels in the initial 5 min of an access session and persist in drinking beyond that period.

Our frontloading data from the multiple regression, where predictors were group-centered by sex, provide some explanation for our observed sex differences. Females reached higher breakpoints than males, and displayed a higher binge and lower onset skew pattern of drinking, suggesting that frontloading differences, though unrelated to overall sex differences in consumption, may underlie sex differences in addiction-like phenotypes. Indeed, higher bingeing and lower onset skew were also both predictive of more persistent extinction pressing, as observed in females. In contrast, onset skew did positively predict resistance to footshock, and this may be reflective of highly concentrated pressing being more impulsive or less controlled and more difficult to suppress in the face of aversive consequences.

Conflict preference sex-dependently predicts addiction-like phenotypes

We hypothesized a sex-dependent relationship between conflict preference and addiction-like drinking, and our data provide partial support for the hypothesis. We found a baseline difference in conflict preference between males and females, with males approaching the conflict cue more than females, consistent with other studies on risky decision-making, where males demonstrate more risky choices favoring larger rewards even when they are coupled with unfavourable conditions such as low probability, aversive outcomes (e.g., shock) or delay.85 Furthermore, we found that conflict preference negatively predicted ethanol-seeking in extinction (faster extinction) and positively predicted compulsive ethanol drinking in males only. The observed increase in conflict preference in the males may generally represent greater weight being placed on positively reinforcing outcomes as reported previously in value-based decision-making tasks11, and could explain their ethanol-seeking being resistant to footshock, as they place greater value on the ethanol outcome, even in the face of aversive outcomes. Furthermore, since conflict-approaching males are more focused on the ethanol outcome, the absence of reward during extinction is more impactful for them, leading to rapid extinction. This interpretation is consistent with the finding that alcohol-preferring P rats, who find alcohol more rewarding than Wistars, demonstrate both greater persistence in acquiring quinine-adulterated ethanol and a faster reduction of ethanol-seeking under extinction conditions37,86. Our findings, and this reward-based interpretation is also consistent with recent work in human men, suggesting that overvaluing incentives (and specifically not underestimating aversive outcomes) predicted meeting more criteria for AUD in the future87.

In females, while we had previously observed a negative relationship between conflict preference and homecage ethanol intake26, our models here detected no influence of conflict preference on any drinking measures in females. Importantly, our previous finding in females was based on free, unconstrained drinking in a homecage IA2BC paradigm26, while the current study required goal-directed, effortful lever pressing from rats to acquire ethanol. It is possible that this additional cost of lever pressing may have masked the influence of conflict avoidance on drinking behavior in females, despite females showing increased motivational drive to consume alcohol, as seen in the overall ethanol consumption levels and higher BEC. This account, however, does not fully align with the present observation that females also reached significantly higher breakpoints in the progressive ratio task, which could reflect not only their motivation for ethanol, but also their willingness to tolerate increasing costs. Further work is clearly warranted to determine how differences in free vs effortful responding for ethanol may be related to conflict preference in females.

Conclusion

In this study, we identified sex differences in how conflict preference predicts future addiction-like measures of drinking, and how potential sex differences in value-based decision making may underlie different addiction-related phenotypes. We had hypothesised that conflict approach bias would predict addiction-like phenotype in males more robustly than in females, and we found the influence of conflict preference to depend on sex in predicting both persistent ethanol seeking in extinction and continued ethanol taking in the face of footshock, building on our previous findings demonstrating a sex-dependent relationship between approach-avoidance conflict and voluntary homecage ethanol drinking26. Surprisingly, the direction of our observed effects differed by addiction-like criterion, with greater conflict approach in males related to reduced responding in extinction but elevated consumption despite footshock. Following recent work in humans showing the utility of value-based decision-making in predicting the number of AUD criteria males meet in the future87, our work suggests the same predictive relationship may not hold for females. Additionally, more work can be done in the preclinical realm to operationalize different criteria for AUD to make better and more reliable models and to continue to appreciate measures of alcohol-related behavior beyond simple indices of intake. The present study suggests that risk factors for the development of addiction-like behaviors differ by sex, which may have predictive relevance for humans, and also highlight the potential to individualize treatment approaches after the development of addiction.

Beyond our primary hypothesis, we identified a more nuanced relationship between frontloading, which is often used as a model for binge-like behavior in humans, and motivation than previously suggested43. In the present study, frontloading measures of intake positively predicted addiction-like phenotypes while frontloading measures of skew negatively predicted addiction-like phenotypes. Our findings suggest that examining individual differences using continuous indices of frontloading across two dimensions (i.e., bingeing and onset skew) may offer valuable insights into addiction vulnerability. Females overall displayed greater ethanol consumption and vulnerability to addiction-like criteria than males, and this profile was associated with a high binge, low onset skew phenotype. Future work can examine addiction-related criteria in more and less effortful contexts to further probe how sex differences in value-based decision making may underly sex differences in drinking.