Long Evans rats were bred in our facility. For the Experiment 1 data, there were 50 rats (21 male, 29 female). For the Experiment 2 data, there were 103 rats (48 male, 55 female). Rats were weaned on PND 21. Following weaning, animals were individually housed and maintained on a 12-hour reverse light-dark cycle with lights off at 07:00 am in a temperature and humidity-controlled room. Water was available ad libitum throughout the experiment and food was available ad libitum until the beginning of the food restriction period. At PND66, the rats were food-restricted to 85% of their initial free-feeding weight by daily feedings with a minimum of 5 g of food chow per day. Once rats reached their 85% target weights, the target body weight increased by 1 g/day for males or 0.25 g/day for females for the remainder of the experiment, such that the rats’ target weights gradually increased by 7 g/week or 1.75 g/week. The different feeding regimens in male and female rats are meant to cause similar leverpress rates in the two sexes (determined from our lab’s pilot data), and is consistent with our published research using the devaluation task [40].

All procedures and animal care were approved by and conducted in accordance with the Kansas State University Institutional Animal Care and Use Committee guidelines (approvals #4345 and 4422) and were conducted in accordance with the U.S. National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. All rats were euthanized with a lethal dose of Fatal-Plus (pentobarbital solution; #2831, 2846, and 2868; Vortech Pharmaceuticals, Dearborn, MI, USA) followed by decapitation as a secondary method of confirming death after the completion of the experiments, consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.

The behavioral training occurred in 12 operant chambers (Med Associates, St Albans, VT, USA). Each chamber had a pellet dispenser that was used to deliver 45-mg precision pellets (catalog #1811155; TestDiet, Richmond, IN, USA), 45-mg chocolate-flavored sucrose pellets (catalog #1811256; TestDiet, Richmond, IN, USA), or 45-mg grain pellets (catalog #1812127; TestDiet, Richmond, IN, USA). The particular pellet that was delivered depended upon the phase of the experiment. Each chamber was equipped with two retractable levers on either side of the food-cup at approximately one-third of the chamber’s total height. A white cue-light was located above each of the levers. A red house-light was located on the back wall in the top–center. The equipment was controlled by (and all responses were recorded by) a Dell OptiPlex computer (Dell Inc., Round Rock, TX, USA) with Med-PC for Windows (Med Associates, St Albans, VT, USA).

Behavioral training occurred after rats had acclimated to the restricted feeding conditions. This training occurred during the dark phase of the light:dark cycle. Rats were weighed and received their daily feeding after each day’s training session.

This experiment utilized the same training and testing procedures used and described in detail by Fisher, Pajser, and Pickens [40]. In brief, the rats were trained with three magazine training sessions (one with each of the three types of food pellets- chocolate, grain, and precision). The magazine training sessions occurred once-daily for 40 min. During each session, one of the pellets was delivered every 125 seconds. The rats were then trained to leverpress in two fixed-interval-1 training sessions during each day for two days (with afternoon sessions occurring 2–4 hours after the morning sessions). In these sessions, rats could earn up to 50 grain pellets and leverpresses could earn a pellet each second under the fixed-interval-1 schedule. Each lever was trained in a separate session for a maximum of 50 pellets or 60 minutes per session, with training for each lever in the morning on one day and in the afternoon on the other day.

After this pretraining was complete, the rats were trained in 4 cued-trial operant sessions, given once daily. The rats were trained with two sessions for each of the two lever-cue-reinforcer relationship. During the first and the fourth training session, responses to the left lever (with illumination of a steady white cue-light above the lever) resulted in precision pellets. During the second and the third training session, responses to the right lever (with flashing [2 Hz] illumination of white cue-light above it) resulted in chocolate pellets. Each cued-trial training session lasted for 40 min and contained 40 trials. During each 40-sec trial, leverpresses could earn two pellets, with one pellet available at a variable time during the first 20 sec and pellet available at a variable time during the last 20 sec (a modified VI20 reinforcement schedule during the trials). Rats could earn a maximum of 80 pellets during each session. During the inter-trial intervals (13–19 sec with a mean of 20 sec), the levers were retracted without any illumination of the cue-lights.

Following cued trial training, a choice test was administered. In the operant chambers, rats received a 1-h satiation session with access to 30 g of either chocolate or precision pellets (identity counterbalanced) presented in ceramic bowls. Fifteen minutes after the satiation period, rats received a 12-min choice test with twelve 40-sec trials. During each trial, both levers with cue-lights were presented and responding was measured. In the Cue Normal condition, the cue-lights above the levers were in the same position as during training (e.g., steady cue-light above the left lever). Responding should be decreased for the cue-light+lever compound that earns the devalued reinforcer if rats devalue properly. For the Cue Switched condition, the cue-lights above the levers were in the opposite position as during training, such that the flashing light was above the left lever and the steady light was above the right lever. Our previous research suggests that neurobiologically intact rats decrease responses made to the lever in the fixed location that previously earned the devalued/sated food rather than the lever below the cue-light associated with the sated food under these conditions [40]. However, our preliminary research suggests that PL lesions lead to impaired devaluation with a trend towards lower responding to the lever below the cue-light associated with the sated food, even though this lever-location is associated with the non-sated food (unpublished data). Lever presses did not earn pellets during the test to ensure the rats used their memory representations of the cue-lever combinations to guide behavior. Next, rats received two cued-trial retraining sessions, one with each lever+cue-light-food compound (with the original configurations used during training), to return lever pressing rates to baseline. Rats then completed another choice test with the opposite pellet devalued during the satiation period and with the same cue-light-+lever combinations as in the first test (so Cue Switched groups once again had mismatched cue-light+lever compounds between training and test). Testing with both pellets sated controls for any differences in preference for them, as we often observe higher lever-responding for the precision pellets and higher consumption of them during the satiation period (see [67] for an example). Any preference should be averaged out by testing with both pellets sated. No rats were excluded for low responding in the choice tests. However, in Experiment 1, one rat was excluded because of a jammed feeder that led to extinction of the lever-press response and one rat was excluded as an experimental error led to them not being tested in a second choice test.

We have traditionally assessed devaluation effects in our previous research by assessing the number of leverpresses, but an alternative measure of the devaluation effect can be calculated by assessing the ratio of responding on the lever that previously earned the nondevalued food to responding on the lever that previously earned the devalued food. We calculated this using the formula (with the “nondevalued” and “devalued” lever designated based on the levers’ fixed spatial location):

$$\begin{array}{c}\hfill \text{Preference ratio}=\hfill \\ \hfill \frac{(\text{Responses on the nondevalued lever – Responses on the devalued lever})}{(\text{Responses on the nondevalued lever + responses on the devalued lever})}\hfill \end{array}$$

This measure normalizes the change in responding based on the overall level of responding. For example, using this measure, a low responding rat that presses 4 times on the nondevalued lever and 2 times on the devalued lever would have a preference ratio equivalent to another rat that presses 30 times on the nondevalued lever and 15 times on the nondevalued lever. In this case, exclusive responding on the nondevalued lever results in a preference ratio of 1.0, exclusive responding on the devalued lever results in a preference ratio of –1.0, and equal responding on the 2 levers results in a preference ratio of 0.0.

3.3.1 Preference Ratios in Experiment 1

We analyzed the test results of Experiment 1 (Fig. 4A,B) using a mixed-factor ANOVA with the between-subjects factors of Treatment (Amphetamine and Saline), Sex (Male and Female), and Test Type (Cue Normal and Cue Switched). This analysis found a no significant main effects or interactions (all p $>$ 0.32). We then examined whether the overall pattern using this measure supported a conclusion that goal-directed action was being expressed using a single-sample t-test. As there were no differences between groups, we compared the preference ratio across the entire sample to 0 (the score that would indicated that there was no preference). We found that the preference score for the sample was significantly different from 0: t(46) = 5.3, p 0.001. These results indicate that there was a significant devaluation effect (based on the lever-location) in Experiment 1 when the scores were normalized for the overall level of responding and that there was no evidence that the magnitude of this devaluation effect using this measure differed across the different groups.

Fig. 4.

Preference ratio data (mean + SEM) in each group during the devaluation test from Experiments 1 and 2. (A) Preference ratio in Experiment 1 averaging across sex. (B) Preference ratio in Experiment 1 separated by sex. (C) Preference ratio in Experiment 2 averaging across sex. (D) Preference ratio in Experiment 2 separated by sex. Pref, preference; F, female; M, male; Normal, behavior in the Cue Normal test; Switch, behavior in the Cue Switched test.

Our results were not informative about the possibility of compensation between strategies after drug exposure. We designed and conducted the experiments specifically to examine the possibility of impaired devaluation in the Cue Switched condition and not in the Cue Normal condition, which would indicate compensation between strategies used to guide goal-directed behavior. The failure to find impaired devaluation in the Cue Switched condition in the drug-exposed groups means our results are uninformative about whether compensation after drug exposure can help to maintain goal-directed behavior. However, our results largely replicate our previous findings about the relative role of the two types of strategies that can guide behavior. In both experiments, we replicated our previous pattern in which rats tested under the Cue Switched conditions primarily guide their responding according to the fixed location of the lever rather than based on the cue-light’s location [40]. Visual inspection of the devaluation effect in Experiment 2 using the leverpress/trial measure (Fig. 3B,C) suggested that the devaluation effect may have been smaller in the Cue Switched condition, but this was not supported by a Lever X Test Type interaction in the analysis of either the leverpress/trial measure or the preference ratio measure in Experiment 2. This visual pattern was not observed in Experiment 1. Additional research is necessary to assess whether different training conditions could lead to goal-directed behavior that is guided by cue-lights in neurobiologically intact rats.

Our finding that amphetamine exposure did not impair goal-directed behavior contrasts with several prior demonstrations that psychostimulant exposure can lead to altered PL and/or lOFC function and goal-directed behavior impairments in devaluation tasks dependent on PL or lOFC function. Prior research has shown that psychostimulant exposure (through forced exposure or self-administration) alters the activity of PL and/or lOFC. For example, repeated amphetamine/cocaine injections or methamphetamine self-administration increases excitability and/or firing rate in PL days or weeks later [68, 69, 70, 71, 72]. Previous cocaine exposure also disrupts selective lOFC firing to cues that predict specific outcomes in a discrimination task [73]. There is also evidence for impairments in goal-directed behavior weeks after the final cocaine exposure in a devaluation task requiring a cue-based strategy [56], although the impairment could be explained by training and testing in a drug-paired environment rather than neurotoxic dysfunction [58, 59]. Likewise, prior research has demonstrated behavioral impairments weeks after the final cocaine, amphetamine, or methamphetamine exposure in a devaluation task requiring a lever-location strategy [51, 52, 53, 54, 55], although the impairment may be caused by faster habit learning rather than a failure of the neurobiological substrates of goal-directed behavior [57].

Our results may be congruent with the prior literature showing goal-directed behavior impairments in the devaluation task after prior psychostimulant exposure if these prior impairments were due to accelerated habit-formation (in tasks prone to formation of habits after extended training) or training/testing in drug-paired contexts rather than a stable neurotoxicity-induced dysfunction of the brain areas needed for goal-directed behavior. As noted above, prior research has demonstrated impairments in goal-directed behavior weeks after the final cocaine exposure in devaluation tasks requiring a cue-based strategy [56], but this experiment paired the behavioral training and testing environment with cocaine. Previous research has also shown that behavioral contexts associated with the psychostimulant methamphetamine lead to devaluation impairments that are not observed in the same rats when tested in a different context [58], so the cocaine-induced impairment in the cue-based task may be due to an interference effect from the cocaine-paired training and testing context. Many of the other demonstrations of impaired goal-directed behavior in devaluation tasks where behavior was guided by lever-location after cocaine or amphetamine exposure have used task versions with training with one lever-reinforcer relationship [51, 52, 54, 55] or a second response-reinforcer relationship was trained only after the cessation of training for the other response-reinforcer relationship [53]. Extended training can lead to habit-based devaluation insensitivity with variable interval schedules and/or extended training if only one response-reinforcer relationship is trained or if response-reinforcer relationships are trained in different contexts [1, 27], but goal-directed behavior in operant responding in rodents occurs if multiple response-reinforcer relationships are trained in the same context [74].

Our results mirror a pattern observed in multiple-response/multiple-reinforcer operant devaluation (conditions that don’t tend to lead to habit-formation) after exposure to other drugs of abuse. Prior exposure to subchronic cocaine or methamphetamine [57, 58, 75] or a neurotoxic single-day regimen of methamphetamine [76] do not cause a generalized goal-directed behavior impairment in devaluation tasks where two different response options lead to separate rewards, possibly because these procedures are resistant to the formation of habits [74]. In these cases, cues associated with a drug can specifically interfere with goal-directed behavior in the presence of those cues [58, 75], but no impairment is seen when these cues are absent.

However, there are several methodological considerations. First, the timing of the amphetamine exposure may have affected our results. While previous research suggests that there may be more significant long-term behavioral and neurobiological effects of drug exposure if this exposure occurs during adolescence rather than adulthood (as reviewed in [50]), it is possible that amphetamine exposure during adolescence had less of a long-term effect on behavior than if it were given during adulthood. Relatedly, it is possible that we could have found an impairment in devaluation if we had trained and tested the rats with a shorter interval between amphetamine exposure and training/testing. Second, it is possible that higher doses of amphetamine could lead to impaired devaluation. We based our injection regimen and time of exposure on a previous study that showed that 3 mg/kg injections of amphetamine every-other day from PND 27–45 led to decreased bursting in PL in response to saline injections during adulthood [60]. Our 3 mg/kg dose was also higher than the 2 mg/kg dose that was previously shown to impair devaluation in a task that predisposes rats towards habit learning if rats are trained for enough sessions (in a previous experiment with 7 once daily injections for 7 days during adulthood) [54]. However, we did not assess the effects of amphetamine exposure at a range of doses and it is possible that higher doses would lead to devaluation impairments in tasks that do not predispose rats to habit learning if they are extensively trained. Third, while our predicted impairments in goal-directed behavior did not depend upon the presence of sensitization to the effects of psychostimulants, several of the previous reports of psychostimulant exposure impairing devaluation also showed that the impaired rats exhibited locomotor sensitization to the effect of the psychostimulant [54, 56]. As we did not assess locomotor sensitization, it is unclear whether our amphetamine exposure regimen led to sensitization. However, the level of sensitization did not correlate with the level of devaluation impairment in the previous experiments that assessed it [54, 56], so it is also unclear whether psychomotor sensitization is required for devaluation impairments. Finally, it is possible that the timing of our injections (specifically- during the dark phase of the light-dark cycle) may have affected our results, as there is prior evidence for the light-dark cycle affecting the sensitivity to, pharmacokinetics of, and/or intake of a variety of drugs of abuse (for example, see [77, 78, 79, 80]). We gave injections during the dark phase to rats on a reverse light-dark cycle, while most of the previous experiment examining long-term effects of psychostimulants on devaluation and that reported their lighting conditions reported maintaining rats on a regular light cycle (with injections presumably occurring during the day) or specified injecting during the light phase [53, 54, 56, 58, 81] (but see [55]).

While we found relatively clear evidence that the particular amphetamine exposure regimen and washout period did not lead to devaluation impairment in our task, additional research is needed. Specifically, future research should (1) determine if our amphetamine exposure regimen would lead to accelerated habit formation in tasks that predispose rats towards habit learning if rats are extensively trained, (2) determine if this exposure regimen would impair behavior in the devaluation if given during adulthood, (3) determine if other exposure regimens such as higher doses or shorter washout periods might lead to a goal-directed behavior impairment, (4) determine if our amphetamine exposure led to an impairment in lOFC function that was obscured by intact PL function or whether it did not impair PL or lOFC function, and (5) determine if this exposure regimen leads to locomotor sensitization.

Our finding that ketamine exposure did not impair goal-directed behavior contrasts with several prior demonstrations that dissociative anesthetic exposure can lead to altered PL and/or lOFC function and impairments in tasks dependent on PL or lOFC function, although the prior literature on devaluation tasks is relatively sparse. One previous experiment found that acute injections of the non-competitive N-methyl-D-aspartate (NMDA) receptor (+)-5-methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine maleate (MK-801) between the satiation period and the leverpress devaluation test impaired the selective decrease in leverpressing for the reward in a devaluation task in which only one response-reward relationship had been trained (with decreased leverpressing regardless of whether the operant reinforcer or another food was sated) [82]. However, this experiment examined acute effects with pretest injections rather than long-term effects in a drug-free state. In addition, the NMDA receptor antagonist used (MK-801 vs. ketamine) and the type of devaluation task (single lever-reward free operant task vs. two response-cue-reward relationships) differed from the parameters used in our study. Therefore, it is unclear which factor or factors distinguish our results from this previous one. Conversely, our lab previously examined long-term effects of a ketamine regimen with higher doses and fewer injections (3 once-daily injections of a higher sub-anesthetic 50 mg/kg or anesthetic 100 mg/kg dose) and found no effects of ketamine exposure on goal-directed behavior in the same task used here with a Cue Normal test [83]. However, there was no Cue Switched test given, so it is unclear whether this represents intact function of PL, lOFC, or both, as goal-directed behavior in the Cue Normal task is intact if only PL or only lOFC is lesioned [44], and the fewer number of exposures may have led to a less pronounced effect than long-term exposure. Therefore, there is little to no data on what would be expected for long-term effects of our ketamine exposure regimen in the current experiment.

It is unclear whether our ketamine exposure regimen failed to alter prefrontal cortex (PFC) function or whether any impairments may have occurred but did not last until our training/testing period. Prior literature suggests that there are impairments in the reversal learning task (which is dependent on lOFC [62, 63, 64]) after subchronic exposure to ketamine or the NMDA receptor antagonist drug phencyclidine (PCP) under some conditions [61, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93] and no impairment under other conditions [65, 94, 95, 96, 97, 98, 99, 100]. Likewise, prior literature suggests that there are impairments in extradimensional shifting in the attentional set-shifting or strategy set-shifting tasks (which is dependent on PL [66, 101, 102]) after subchronic exposure to PCP or ketamine [61, 65, 85, 91, 95, 97, 98, 99, 100]. There were relatively uniform regimens of exposure and washout in these previous studies, with the vast majority of experiments giving dissociative anesthetics during adulthood and having testing between 3 and 14 days later (compared with the much wider range in age of exposure and washout periods ranging from 24 h to 6 weeks for psychostimulant exposure and testing). While this simplifies the prior literature, it may limit the generalizability of these findings to make predictions about our experiment.

There are several methodological considerations for the interpretation of our findings. First, our devaluation effect was smaller in Experiment 2 than in Experiment 1. Although all 8 groups (across each Male vs. Female X Ketamine vs. Saline X Cue Normal vs. Cue Switched condition) exhibited numerically lower responding to the devalued lever than to the nondevalued lever, it is possible that this weaker overall devaluation effect may have made it more difficult to detect any potential decrease in the magnitude of the devaluation effect after ketamine. This was particularly evident in visual inspection of Saline group rats given Cue Switched conditions, although there was no significant Test Type X Lever interaction. Second, it is possible that higher doses of ketamine could lead to impaired devaluation. We based our injection regimen on dosing regimens previously shown to lead to behavioral impairments consistent with dysfunction of PL and/or lOFC [61, 65]. However, we did not assess the effects of ketamine exposure at a range of doses or lengths of exposure, and it is possible that higher doses or more extensive exposure are required to lead to devaluation impairments. Third, the timing of the ketamine exposure may have affected our results. As the experiments reviewed in the previous paragraph were performed on rodents during or near adulthood and they all began behavioral testing within ~2 weeks or less after the final PCP or ketamine injection, it is unclear how relevant they are for predicting PL and lOFC function more than 5 weeks after ketamine exposure in adolescent rats. While previous research suggests that there may be more significant long-term behavioral and neurobiological effects of drug exposure if this exposure occurs during adolescence, rather than adulthood (as reviewed in [50]), it is possible that ketamine exposure during adolescence had less of a long-term effect on behavior than if it were given during adulthood. In addition, previous research has found that an identical once-daily PCP exposure led to opposite effects on activity regulated cytoskeletal-associated protein (Arc) mRNA in PL and lOFC 3 days after the final injection in rats of different ages, with decreased mRNA in adolescent rats and increased mRNA if given during early adulthood [103]. Alternatively, it is possible that adolescent ketamine exposure did lead to altered function of PL and/or lOFC for several weeks after exposure, but the more extended washout period (more than 5 weeks from the final injection until cued-trial training began) in the current experiment led to a recovery over time.

While we found no evidence that the particular ketamine exposure regimen and washout period led to devaluation impairment in our task, additional research is needed. Specifically, future research should (1) determine if our ketamine exposure regimen would lead to accelerated habit formation in tasks that predispose rats towards habit learning if rats are extensively trained, (2) determine if this exposure regimen would impair behavior in the devaluation if given during adulthood, (3) determine if other exposure regimens such as higher doses or shorter washout periods might lead to a goal-directed behavior impairment, and (4) determine if our ketamine exposure regimen led to an impairment in lOFC function that was obscured by intact PL function or whether it did not impair PL or lOFC function.