Executive functioning-specific behavioral impairments in a rat model of human third trimester binge drinking implicate prefrontal-thalamo-hippocampal circuitry in Fetal Alcohol Spectrum Disorders

Abstract

Individuals diagnosed with Fetal Alcohol Spectrum Disorders (FASD) often display behavioral impairments in executive functioning (EF). Specifically, the domains of working memory, inhibition, and set shifting are frequently impacted by prenatal alcohol exposure. Coordination between prefrontal cortex and hippocampus appear to be essential for these domains of executive functioning. The current study uses a rodent model of human third-trimester binge drinking to identify the extent of persistent executive functioning deficits following developmental alcohol by using a behavioral battery of hippocampus- and prefrontal cortex-dependent behavioral assays in adulthood. Alcohol added to milk formula was administered to Long Evans rat pups on postnatal days 4–9 (5.25 g/kg/day of ethanol; intragastric intubation), a period when rodent brain development undergoes comparable processes to human third-trimester neurodevelopment. Procedural control animals underwent sham intubation, without administration of any liquids (i.e., alcohol, milk solution). In adulthood, male rats were run on a battery of behavioral assays: novel object recognition, object-in-place associative memory, spontaneous alternation, and behavioral flexibility tasks. Alcohol-exposed rats demonstrated behavioral impairment in object-in-place preference and performed worse when the rule was switched on a plus maze task. All rats showed similar levels of novel object recognition, spontaneous alternation, discrimination learning, and reversal learning, suggesting alcohol-induced behavioral alterations are selective to executive functioning domains of spatial working memory and set-shifting in this widely-utilized rodent model. These specific behavioral alterations support the hypothesis that behavioral impairments in EF following prenatal alcohol exposure are caused by distributed damage to the prefrontal-thalamo-hippocampal circuit consisting of the medial prefrontal cortex, thalamic nucleus reuniens, and CA1 of hippocampus.

1. Introduction

Prenatal alcohol exposure often results in diagnosis of a Fetal Alcohol Spectrum Disorder (FASD) [1]. The prevalence of FASD has a heterogeneous distribution throughout the world, but is commonly believed to be between 1–10% in many countries [2], with alcohol exposure during late pregnancy expected to be around 8% in some parts of the United States [3]. Despite this already high rate of diagnosis, there is still a significant rate of missed diagnosis (as high as 80%) and misdiagnosis [4], suggesting that FASD may pose a greater public health concern than previously recognized. Individuals diagnosed with FASD tend to display behavioral disruptions on complex cognitive tasks such as working memory, inhibition, and set-shifting [5–7], which access executive function [8].

A number of animal models have been developed to assess the impact of alcohol exposure at different developmental time points, and in different doses [reviewed in 9, 10]. One particular model, administering ethanol in a milk substitute to rats during the first two postnatal weeks of life via intragastric intubation [e.g., 11, 12], has gained popularity for its ability to precisely control the dose of ethanol administered to the pup. The first two postnatal weeks of life in rat also model a period in nervous system development referred to as the “brain growth spurt” [13], which occurs during third trimester of humans, making the above-mentioned model ideal to study the impact of alcohol exposure during late pregnancy in humans.

The brain growth spurt model of developmental alcohol exposure has consistently produced damage to structures critical for learning and memory, such as the hippocampus (HPC) [14–16] and medial prefrontal cortex (mPFC) [15, 17–19]. Although a body of literature has demonstrated impairment in traditionally hippocampus-dependent tasks in rodent models of human third-trimester alcohol exposure [14, 20, 21], emergent literature suggests that activity in mPFC may be interrupted even when HPC activity appears normal [22].

mPFC and HPC form a behaviorally relevant circuit which depends on strong reciprocal connectivity with nucleus reuniens (Re) of the ventral midline thalamus [23], giving rise to the hypothesis that the dysfunction and damage observed in mPFC or HPC in FASD may actually be caused by dysfunctional input from Re throughout life. Despite the necessity of Re to synchronize activity between mPFC and HPC during complex behaviors such as spatial working memory [24], the impact of developmental alcohol exposure on this nucleus has only recently been examined. Using the same rat model of third-trimester alcohol exposure as Heroux, Robinson-Drummer [22], Gursky, Savage [25] were the first to examine Re integrity in adulthood, observing a significant permanent loss of neurons within Re. No damage was observed in the neighboring rhomboid nucleus of midline thalamus, suggesting specific damage to Re, likely due to its high level of interconnectivity with mPFC and HPC. The most probable cause of this neuronal loss is the apoptotic cell death occurring just hours after the alcohol exposure occurs [26], resulting in long-lasting Re dysfunction. Permanent Re neuron loss caused by developmental alcohol exposure may contribute to underdevelopment of HPC and PFC, or spontaneous circuit dysfunction during HPC- and PFC-dependent behaviors. Using a battery of behavioral tests that depend on PFC (reversal learning and rule switching), or HPC and PFC (object in place, spontaneous alternation), the current study examines whether dysfunction of HPC and/or PFC that accounts for behavioral impairments in FASD, or whether the constellation of behavioral impairments can more parsimoniously be explained by damage to Re.

Due to the complexity of behavior that is supported by the mPFC-Re-HPC circuit, we hypothesized that early postnatal alcohol exposure would cause deficits in executive functioning — but not non-associative memory or discrimination learning — in adult male rats. We ran rats on a behavioral battery consisting of a novel object recognition task (performance control for object-in-place), object-in-place task, spontaneous alternation task, and behavioral flexibility task (including discrimination learning, reversal learning, and rule switching). We hypothesized that alcohol-exposed rats would be impaired on tasks that require executive functioning domains that are commonly altered in FASD (i.e., spatial working memory, rule switching), but not on tasks with reduced complexity such as object recognition, spontaneous alternation, and discrimination learning.

2. Materials and methods

2.1. Experimental Subjects

All animal procedures were performed in accordance with the University of Delaware animal care committee’s regulations. A total of 28 male Long Evans rats from 9 litters were used in the current study. All animals were obtained from in-house breeding colony at the University of Delaware and housed with a 12-hour light-dark cycle (lights on at 9 AM). Food and water were administered ad libitum. On postnatal day (PD) 3, litters were culled to 8 pups (4 males and 4 females, where possible). Pups were cross-fostered on PD 3 when necessary to establish consistent litter size. Pups were also administered subcutaneous injections of small amount of black India ink in the pad of the paw to provide a unique identifier for each animal. Pups within a litter were randomly placed into one of two experimental groups: alcohol exposed (AE) or sham-intubated (SI). Each litter consisted of 4 AE and 4 SI animals. Only male pups from these litters were used for the current study; females were used for a different study [25].

2.2. Experimental Manipulation (PD 4–9 Ethanol Administration) and Weaning

A timeline summarizing the current study can be found in Fig 1. On PD 4–9, all animals were weighed daily at 9AM. AE animals were administered 5.25 g/kg/day of 11.9% v/v ethanol in a milk substitute via intragastric intubation over 2 doses, 2 hours apart. AE animals also received 1 daily supplemental dose of milk substitute (2 supplemental doses on PD 4) 2 hours after the second ethanol administration to prevent confounding weight differences from insufficient feeding during intoxication. Blood samples were collected from the tail vein on PD 4 at 90 minutes following the second ethanol administration, as blood alcohol content (BAC) is at peak at this time relative to administration, and reaches its highest peak on PD4 (assuming similar doses of AE on each day) [27]. Collected blood samples were centrifuged at 15,000 × g for 25 minutes at 4 °C; plasma was removed from supernatant and stored at −20°C until analysis. BAC was analyzed using an Analox GL5 Alcohol Analyzer (Analox Instruments, Boston, MA). Mean (±SEM) peak BAC for the current study was 360.54 ± 39.47 mg/dL (range: 58.2 to 511.5 mg/dL).

Fig 1: Experimental timeline for all manipulations.

While litters initially contained both male and female rat pups, all behavioral assays were performed on male Long Evans rats. The number in parentheses below each behavioral task on the timeline indicates the number of days over which the task occurred.

SI animals underwent identical procedures on PD 4–9 (intubation, blood collection) but did not receive any milk substitute (due to potential weight confounds like those observed by Goodlett and Johnson [28]) or ethanol. All animals were earpunched on PD 9 following the final milk dose, for litter identification purposes.

On PD 23, all animals were weaned and subsequently housed in social cages of 2–3 same-sex animals, as is common in studies of developmental alcohol exposure [21, 29].

2.3. Behavioral Apparati

Two distinct behavioral apparati were utilized to assess hippocampus- and prefrontal cortex-dependent behaviors in adulthood: an open field arena and a plus maze. The open field arena (used in novel object recognition and object in place tasks) is identical to the apparatus used by [30], and measures 90 cm × 90 cm × 60 cm. The plus maze (used in spontaneous alternation and behavioral flexibility tasks) consisted of 40 cm × 15 cm painted black PVC arms with 20 cm tall Plexiglas walls and was similar to the apparatus used in [31]. The plus maze also included small ventilated compartments below each goal arm to allow placement of inaccessible rewards below the reward location, preventing the confound of scent-based reward-seeking.

2.4. Behavioral Testing

All rats used in the current experiment underwent all behavioral tests in a fixed order. That order was: Novel Object Recognition, Object in Place, Spontaneous Alternation, then Behavioral Flexibility.

2.4.1. Handling and Arena Habituation

Once male animals reached adulthood (PD 52–72; at least 300g), each animal was handled for 5 min per day for 3 consecutive days. After 3 days of handling by experimenters, animals were habituated to the open field arena (to be used for the novel object recognition and object in place tasks) for 2 days, 10 min on each day. During habituation, animals could freely explore the empty arena, before being returned to the home cage.

2.4.2. Novel Object Recognition (NOR) Testing

Animals were first assayed on a NOR task to confirm that all rats were capable of retaining object identity over a 5 min delay period and show preference for novel objects [32]. The day following the second habituation day, animals were subject to NOR testing as in [33]. NOR testing consisted of a 5 min habituation phase, a 5 min sample phase, and a 5 min probe phase. Each phase would be separated by a 5 min intertrial interval (ITI) when animal was returned to home cage. During ITI, the arena and objects would be cleaned. Habituation phase on testing day was a 5-minute version of the process from the two prior days. During the sample phase, the animal would be placed in the arena with two similar objects and video-recorded using a camcorder (Sony Corporation; DCR-HC28). During the probe phase, the animal would be placed in the arena with one familiar and one novel object and video-recorded. The location of the novel object was counterbalanced within neonatal treatment.

Exploratory behavior was coded by several experienced research assistants blind to postnatal treatment using the criteria in Barker and Warburton [33] and values for each measure were averaged across research assistants to generate a single value of each measure for each animal. A discrimination index was generated for each animal which was defined as the difference between the amount of time spent exploring the novel object and the familiar object divided by the total amount of time exploring both objects (+1 being exclusive novel object exploration, 0 being equal amounts of novel and familiar object exploration, and −1 being exclusive familiar object preference).

2.4.3. Object in Place (OIP) Testing

The day following NOR testing, animals were subject to object in place testing (OIP), similar to the procedures used by Barker and Warburton [33], due the task’s dependence on mPFC and HPC communication. Identical to NOR testing (see section 2.4.2), OIP testing had 5-minute-long habituation, sample, and probe phases, separated by 5-minute ITI. For the sample phase, each animal was placed in the arena with four distinct objects. In the probe phase, the animal was placed in the arena with the same 4 objects, but two of the four objects were swapped locations. The location and identity of the moved objects were counterbalanced within neonatal treatment. Exploratory behavior was coded and interpreted similar to NOR (see section 2.4.2).

2.4.4. Food Restriction

Following completion of NOR and OIP testing, a randomly-selected subset of animals (due to logistical constraints) were individually housed, monitored and weighed daily, food restricted to 85–90 % of ad libidum weight over 7–10 days in preparation for assessment on spontaneous alternation and behavioral flexibility tasks. Animals remained on food restricted diets to maintain 85–90 % ad lib. weight throughout spontaneous alternation, behavioral flexibility pre-training, and behavioral flexibility tasks.

2.4.5. Spontaneous Alternation (SA) Task

After 7 days of food restriction and appropriate weight loss, animals were tested for spontaneous alternation behaviors as in Bobal and Savage [34]. SA was chosen to probe hippocampal function as it is robustly hippocampal-dependent across many manifestations of the task [35]. In brief, animals freely explore a plus maze for 18 minutes while an experimenter coded arm entries in order. An arm entry was defined as all 4 paws of an animal passing onto the PVC plank that constitutes a given arm. The total number of entries were quantified and a percent alternation was determined for each animal as in Bobal and Savage [34].

2.4.6. Plus-Maze Habituation and Pre-training

The two days immediately following SA testing, animals were habituated to foraging for food reward (Bio-Serv Dustless Precision Pellets®, 45 mg, Sucrose, Product # F0023) on the plus maze apparatus. Animals were placed in the center of the plus maze and required to consume a food reward placed in the cup of all 4 arms within 2.5 min over 2 consecutive trials. Following 2 consecutive trials of successful foraging, the animal was returned to its home cage.

Starting on the day following foraging training, animals had 2 consecutive days of pre-training. The plus maze apparatus was converted into a T-maze by placing a weighted divider into one of two opposing “start arms” (see Fig 2). Animals were placed in the start arm opposite the weighted divider and required to forage both open arms (“goal arms”) for 4 consecutive trials per day, over 2 consecutive days. The direction and location of the animal’s first goal arm entry was recorded to determine whether the animal had any location or action preferences. No animals displayed marked location-or first turn-preference (data not shown).

Fig 2: Schematic Diagram illustrating Behavioral Flexibility task progression.

Start arms are labeled with an “S”. Goal arms are labeled with a “G”. The rewarded goal arm is indicated with a shaded arrow, with the curved path illustrating the path from start point (black circle) to the reward.

2.4.7. Behavioral Flexibility Task

Following pre-training, animals began training and testing on the behavioral flexibility task, as outlined in Young and Shapiro [31]. The behavioral flexibility task is designed to assess functionality of prefrontal cortex, with different phases of the task probing functionality of distinct prefrontal subregions [36]. Animals were tested during approximately 6 of 7 days in a given week. The task consists of 5 phases, in order: initial discrimination learning (“Disc”), a first reversal (“Rev1”), a second reversal (“Rev2”), a rule switch (“RS”), and a third, post-RS, reversal (“Rev3”). The rules for the behavioral flexibility task were determined by either the response that the animal would have to perform at the center of the maze (“choice point”) to enter the goal arm containing the reward (egocentric, response-based rule), or the goal arm (allocentric, location-based rule) in which the reward was placed. Each rule (i.e., egocentric or allocentric) had 2 contingencies (egocentric: right or left; allocentric: north or south). Each animal was randomly assigned to an initial discrimination rule (action or location) and contingency (left or right; north or south), which was counterbalanced across postnatal treatments. A schematic representation of all initial contingencies, reversals and rule switches is provided in Fig 2.

Animals were not allowed to self-correct following an unsuccessful trial. In a given phase, an animal underwent “training days” during which the animal was given up to 40 trials to reach a criterion of 6 consecutive correct trials. Every 2 consecutive successful trials, the start arm would switch to the opposing (previously blocked) arm. If an animal failed to reach criterion in 40 trials, the subsequent day would be an additional training day. If the animal achieved criterion within 40 trials, the animal would undergo a “retention day” the following day. On a retention day, the animal would perform 24 trials with the start arm randomly determined by a random number generating algorithm. If the animal performed at an 85% or greater success rate on a retention day, it would progress to the next phase of the task on the following day. If the animal performed below an 85% success rate, it would return to a training day for the same phase on the subsequent day. Animals that could not acquire an initial discrimination within 100 total training day trials were no longer run on the behavioral flexibility task (6 total animals [4 SI, 2 AE] excluded for this reason). One SI animal did not complete Rev3 due to mechanical error in the testing room during behavioral testing. Due to this attrition in our original cohort, an additional cohort was generated using identical methods to appropriately power analyses for the behavioral flexibility task, resulting in the addition of 6 total animals (2 SI, 4 AE), which are included in the 28 total animals mentioned in section “2.1. Experimental Subjects.” The total number of trials, successes, errors, and training days to reach criterion were recorded for each animal on each phase of the task.

2.5. Statistical Analyses

All statistical analyses were performed using RStudio (RStudio, RRID:SCR_000432) with R version 3.4.3 [37] running the “tidyverse” [38], “pwr” [39], and “powerAnalysis “ [40] packages. To appropriately analyze all response variables while accounting for potential influence of extreme values (i.e., outliers), the structure of data was comprehensively characterized prior to statistical inference testing. Prior to any between-group analyses or correlational analyses, measures were analyzed for violations of normality using a Shapiro-Wilks test (α = 0.050). If the distribution of a response variable within either group (SI or AE) violated assumptions of normality (p ≤ .050 on Shapiro-Wilks test), that measure was analyzed using nonparametric tests (Wilcoxon rank sum test for between-groups comparisons using the “wilcox.test” function, Spearman rank correlation using the “spearman” method of the “cor.test” function). Data that did not demonstrate violations of normality (p > 0.050 on Shapiro-Wilks test) were analyzed using either a t-test for equal variances (Base R var.test function, p > 0.050) or a Welch t-test for unequal variances (Base R var.test function, p ≤ 0.050). Correlations that did not violate assumptions of normality were performed as bivariate Pearson’s product-moment correlations. All NOR and OIP behavioral testing was compared to chance exploration values (i.e., discrimination index of 0.0) using the appropriate 1-sample statistic mentioned above, as is typical in experiments that examine these behaviors [33, 41] to determine whether control (SI) animals performed the task, to contextualize whether the experimental (AE) group does or does not perform the behavioral task. Effect sizes for between-group analyses, comparisons to chance value, and paired t-tests were calculated using the ES.t.two, ES.t.one, or ES.t.paired functions, respectively (all from the powerAnalysis package). Power analyses for between-group comparisons, comparisons to chance value, paired t-tests, and correlations were performed using the power.t (type = “two”), power.t (type = “one”), power.t (type = “paired”), or pwr.r.test functions, respectively. Statistics are reported as test statistic, degrees of freedom, p-values, effect size (Cohen’s d or r²), and observed power. A list of tests for normality of all response variables in SI and AE animals can be found in Supplementary Tables 1 and 2, respectively.

3. Results

3.1. Novel object recognition is not impaired following early postnatal alcohol exposure

We found that SI (control) animals displayed novel object preference (DI>0) during both five min of the probe phase (t(10) = 6.412, p < 0.001, Cohen’s d = 1.933, power > 0.999, N_SI=11) and the first min of the probe phase (t(10) = 6.700, p < 0.001, Cohen’s d = 2.020, power > 0.999, N_SI=11). Novel object preference in SI control males was greater during the first minute, compared to the full five minutes, of NOR (t(10) = −2.942, p = 0.015, Cohen’s d = 0.930, power = 0.794, N_SI=11). Since this finding is consistent with the work of others that suggest that the first minute of NOR testing is a better predictor of novelty preference than whole-session preference [41], we only analyzed the first minute of the NOR probe phase for comparison between SI and AE animals.

SI and AE male rats did not differ in the total amount of time spent exploring objects during the sample phase (t(20) = −0.287, p = 0.777, Cohen’s d = 0.122, power = 0.068, N_SI=11, N_AE=11) or 1^st min of probe (t(20) = −0.170, p = 0.866, Cohen’s d = 0.073, power = 0.056, N_SI=11, N_AE=11) of the NOR task. Both SI and AE groups displayed preference for the novel object relative to the familiar object during the first minute of probe (t(10) = 6.700, p < 0.001, Cohen’s d = 2.020, power > 0.999, N_SI=11 and t(10) = 5.541, p < 0.001, Cohen’s d = 1.671, power = 0.999, N_AE=11, respectively) (Fig 3). Novel object preference did not differ between AE and SI males (t(20) = 1.704, p = 0.104, Cohen’s d = 0.726, power = 0.653, N_SI=11, N_AE=11).

Fig 3: Rats with higher BACs on PD4 show lower levels of novel object preference.

SI and AE groups did not differ in the total amount of time exploring objects in either the sample, or first minute of the probe phase (3A). Both SI and AE groups did display preference for a novel object, and discrimination index (DI) between postnatal treatments did not significantly differ (3B). The x-axis in 3B indicates a “chance” DI of 0 (similar amounts of time exploring both the novel and familiar object). Data are presented as bars representing mean ± SEM, with sample size in parentheses either within the panel 3A legend, or on the x-axis in panel 3B. Grey bars represent the SI procedural control group, while white bars represent the AE experimental group. # p ≤ 0.050 relative to chance (DI of 0.0)

3.2. Early postnatal alcohol exposure impairs object-in-place preference

SI male rats displayed significant preference for moved objects during the first minute of the probe phase (W(9) = 47, p = 0.049, Cohen’s d = 0.791, power = 0.607, N_SI=10), but not over 5 min of the probe phase (t(9) = 1.471, p = 0.175, Cohen’s d = 0.465, power = 0.261, N_SI=10). Consistent with findings of others that spatial novelty learning is weaker than object recognition learning, and that spatial novelty learning is most robust during the first 1–2 minutes of the probe phase of the task [41], we only analyzed the first minute of the OIP probe phase to contrast SI and AE animals.

The total amount of time spent exploring objects during the sample phase of the OIP task did not differ between SI and AE groups for 5 min (t(19) = −1.644, p = 0.117, Cohen’s d = −0.718, power = 0.345, N_SI=10, N_AE=11), nor during the 1^st min of OIP probe (W(19) = 44, p = 0.468, Cohen’s d = −0.429, power = 0.154, N_SI=10, N_AE=11). Neither SI nor AE animals displayed biased object preference during the sample phase (t(9) = −0.580, p = 0.576, Cohen’s d = 0.184, power = 0.082, N_SI=10, and t(10) = 0.495, p = 0.631, Cohen’s d = 0.149, power = 0.073, N_AE=11, respectively). The SI group showed significant preference for moved objects (W(9) = 47, p = 0.049, Cohen’s d = 0.791, power = 0.607, N_SI=10) but the AE group did not (t(10) = 0.892, p = 0.393, Cohen’s d = 0.269, power = 0.128, N_AE=11) (Fig 4). Despite different performance relative to chance within-group, the SI and AE groups did not significantly differ in moved object preference (W(19) = 59.5, p = 0.778, Cohen’s d = 0.194, power = 0.071, N_SI=10, N_AE=11). However, BAC following final alcohol administration on PD4 negatively correlated with OIP discrimination index (r(9) = −0.635, p = 0.036, r² = 0.403, power = 0.595, N_AE=11). Thus, animals with the highest BACs displayed the lowest OIP discrimination index. Within the AE group, NOR discrimination index did not significantly correlate with OIP discrimination index (r(9) = 0.571, p = 0.066, r² = 0.326, power = 0.479, N_AE=11). The SI group was not examined for this correlation, as they performed significantly higher than chance levels on both tasks.

Fig 4: PD 4–9 alcohol exposure eliminates moved object preference.

SI and AE groups did not differ in the total amount of time exploring objects in either the sample, or first minute of the probe phase (4A). There was no significant preference for the objects that would be moved during the sample phase, prior to moving (4B). The SI group did display preference for moved objects, while the AE group did not, although the SI and AE groups did not significantly differ in their preference for moved objects (4C). The dotted lines in 4B and 4C indicate a “chance” DI of 0 (similar amounts of time exploring both the novel and familiar object). Blood alcohol content (BAC) at 90 minutes following alcohol administration on PD4 showed a strong negative correlation with discrimination index during the OIP task. Data in 4A, 4B, and 4C are presented as bars representing mean ± SEM, with sample size in parentheses within the panel 4A legend, or on the x-axis in panels 4B and 4C. Grey bars represent the SI procedural control group, while white bars represent the AE experimental group. Individual data points are plotted in 4D, while the dashed line indicates the line-of-best-fit for the linear correlation. #p ≤ 0.050 relative to chance (DI of 0.0), n.s. = p>0.050 relative to chance (DI of 0.0)

3.3. Alcohol-exposed animals do not differ in spontaneous alternation behaviors

A randomly selected subset of animals that were run on NOR/OIP were subsequently run on a spontaneous alternation task (SA). SI (n= 7) and AE (n=6) did not differ on either percent alternations (t(11) = −0.043, p = 0.966, Cohen’s d = 0.024, power = 0.050, N_SI=7, N_AE=6) or total number of arm entries (W(11) = 22, p = 0.942, Cohen’s d = 0.513, power = 0.135, N_SI=7, N_AE=6) (Fig 5). However, BAC following final alcohol administration on PD4 showed a negative correlation with percent alternations (r(4) = −0.873, p = 0.023, r² = 0.761, power = 0.698, N_AE=6). While the percent alternation correlation was significant, the range of values for alternations in AE group (15.91% to 51.85%, N_AE=6) is entirely contained within the range of values for SI group (7.58% to 53.66%, N_SI=7). In contrast, BAC did not correlate with number of arm entries (r(4) = 0.803, p = 0.055, r² = 0.644, power = 0.538, N_AE=6).

Fig 5: PD 4–9 alcohol exposure did not alter alternation or activity patterns beyond control levels in a plus-maze spontaneous alternation task in males.

Alternation score did not significantly differ between AE and SI groups (5A). Total number of arm entries did not significantly differ between AE and SI groups (5B). Bars represent mean ± SEM, with sample sizes in parentheses next to group name on the x-axis. The dotted line in 5A represents chance alternation (9.4%). Grey bars represent the SI procedural control group, while white bars represent the AE experimental group.

3.4. Rule switching is the only component of discrimination learning and behavioral flexibility altered by PD 4–9 alcohol exposure

Among animals that could learn an initial discrimination task in under 100 trials (N_SI=4, N_AE=7), AE male rats displayed increased number of trials to criterion (i.e., took longer to learn), relative to SI male rats, during the rule switch phase of the behavioral flexibility task (t(7.037) = −2.806, p = 0.026, Cohen’s d = 1.340, power = 0.480, N_SI=4, N_AE=7), which was due to an increase in the number of successful trials (t(9) = −2.278, p = 0.049, Cohen’s d = 1.428, power = 0.529, N_SI=4, N_AE=7), but not a change in the number of errors (t(9) = −1.669, p = 0.129, Cohen’s d = 1.046, power = 0.320, N_SI=4, N_AE=7) (Fig 6). There was no difference in the number of training days required to learn the rule switch (W(9) = 7.5, p = 0.211, Cohen’s d = 0.930, power = 0.264, N_SI=4, N_AE=7). On all other phases of the behavioral flexibility task (i.e., not the rule switch), AE and SI male rats’ trials, successes, errors, and training days to criterion did not differ (p’s ≥ 0.071, N_SI=4, N_AE=7).

Fig 6: AE males manifest deficits in behavioral flexibility through impaired rule switching.

AE group required approximately twice the number of trials that the SI group required to learn a rule switch (RS) (6A). This was due to an increase in the number of successful trials during the RS (6B), but no difference in the number of errors (6C). There were no significant differences on any of these measures for the initial discrimination (Disc), first reversal (Rev1), second reversal (Rev2), or post-rule switch reversal (Rev3). Bars represent mean ± SEM, with sample sizes in parentheses within the panel 6A legend. Grey bars represent the SI procedural control group, while white bars represent the AE experimental group. *p ≤ 0.050

Neither of the groups (SI or AE) performed significantly better on the Rule Switch (RS) test than the initial discrimination (DISC), both groups required a similar number of trials to master each of these tasks (p’s = 0.494 [SI] and 0.708 [AE]). In the SI group, there was a trend for few trials to reach criterion on the RS, relative to the DISC (p=0.058), but not by the threshold we laid out in the methods/analyses (p≤0.050). However, our protocol required several reversals prior to the RS, and the SI rats were becoming increasing faster at master the tasks. Thus, the traditional increase in trials on the RS was not observed. As has been noted in past studies using this behavioral task, “As rats became familiar with task switching, their performance improved. Normal rats make few errors after several spatial reversals, even when the rule has been changed to a cue-dependent strategy” [42].

4. Discussion

4.1. Third trimester-equivalent alcohol exposure selectively impairs executive functioning

The current study demonstrates that early postnatal alcohol exposure impairs object-in-place associative memory and rule switching behavior. This selective duo of impairments mirror deficits in two domains of executive functioning often present in individuals diagnosed with FASD: spatial working memory and set-shifting [8], suggesting translatable utility of the model of FASD used in this study to the human condition.

The absence of alterations in discrimination learning and reversal learning (orbitofrontal-dependent; Ghods-Sharifi, Haluk [43]) suggests that this developmental alcohol exposure paradigm causes region-specific patterns of damage within the PFC, rather than impaired function of PFC in whole, or inability to learn or perform basic operant discriminations. Since both control and AE groups were able to perform NOR, and impaired OIP performance in AE animals was not correlated with NOR performance, the observed OIP deficit appears to be caused by a selective disruption of spatial working memory rather than a general memory deficit or impairment in object recognition memory. Unaffected motor behavior (i.e., number of arm entries in SA) suggest that motor performance was also unlikely to account for changes in OIP behavior in the AE group. Similar amounts of time exploring all objects during each phase of NOR and OIP demonstrate that changes in OIP discrimination index are driven by differences in object preference rather than changes in activity or total amounts of exploration.

Due to variability in performance of OIP task, as well as BACs, it is imperative that these results be replicated, with a dual focus on dose-dependence and sex differences (for discussion of the latter, see Section 4.3.). The lack of statistically significant difference between treatment groups (SI vs AE) in the current study is notable, despite the performance of each treatment group relative to chance. It appears that a significant between-group difference would only arise from high peak BACs on PD4 (likely over 400 mg/dL, as seen in Fig. 4D), similar to those achieved in other studies that utilize this exposure paradigm [44–47]. The current study provides a foundation for further research to build upon by resolving this question, and the added ability to confirm the importance of the brain-behavior correlates that we now propose underlie such EF-related outcomes (e.g., connectivity of mPFC-Re-HPC circuitry).

4.2. Proposed role for reuniens damage in shaping mPFC function following developmental alcohol exposure

We observed deficits (i.e., more trials to achieve criterion levels of performance) in rule shifting but no abnormalities in reversal learning, therefore, our data suggest that mPFC is uniquely altered within PFC, as mPFC and OFC play dissociable roles in rule switching and reversal learning, respectively [31, 48]. Considering the critical nature of mPFC in rule switching (mentioned above) and OIP [49], the most parsimonious explanation for our observations would appear to be impaired mPFC circuit function, as suggested by Heroux, Robinson-Drummer [22]. We observed a specific increase in the number of successful trials required to reach criterion on the rule switch, while no increase in the number of errors was observed, a pattern inverse to lesion of mPFC, which increases the number of errors, not successful trials when rule switching [48]. In light of recent evidence indicating that lesions to Re impair the ability of mPFC to maintain learning-related increases in mature spines [50] and that developmental alcohol exposure permanently reduces neuron number in Re [25], we propose that impaired mPFC function in rodent models of FASD are likely driven by dysfunction of mPFC-Re-HPC circuitry, rather than extensive pathology directly to PFC.

Ventral midline thalamus activity is necessary for mPFC-HPC synchronization, and inhibition of this area via direct muscimol infusion results in reductions in mPFC-HPC coherence and impaired performance on the (mPFC- and HPC-dependent) delay-non-match-to-place task [24]. Alcohol-induced neuron loss selective to Re [25] likely impairs mPFC-Re-HPC connectivity as the major efferent projections of Re neurons are mPFC, HPC, and ”limbic” structures [51]; however, this hypothesis has yet to be tested. A small subpopulation of Re projection neurons (<10%) that collaterally project to mPFC and HPC [52], which has been hypothesized to be critical for mPFC-HPC synchronization, may constitute an especially vulnerable target for developmental alcohol exposure, as neurons undergoing synaptogenesis are highly vulnerable to developmental alcohol exposure-induced cell death [53]. While no research to date has examined the relationship between damage to mPFC, Re, and HPC simultaneously in FASD, examination of this circuit comprehensively is overdue given the importance of mPFC-Re-HPC development in EF.

While the current study represents a substantial contribution to understanding the consequences of AE during the last trimester of human prenatal development (by using an early postnatal AE paradigm in rodents), studies of prenatal AE in rodents (comparable to first and second trimester in humans [13]) have examined EF more extensively. Vulnerability of PFC to AE seems to be highly subregion-specific in both models of late gestational AE (postnatal in rodents) [19] as well as models of first and second trimester AE (prenatal in rodents) [54, 55], and there seem to be critical differences between deficits in behavioral flexibility observed following prenatal and postnatal AE in rodents. While there is a consensus that behavioral flexibility is impaired in rodent models of FASD, the type and degree of inflexible behavioral tendencies seems to be dependent on the specific task used, method and timing of AE, and age at which animals are tested [56–58]. The current study uses a task for which the functional correlates are thoroughly studied [31, 42, 59]. Thus, future replications of the current behavioral findings are warranted to identify and manipulate specific targeted mechanisms by which early postnatal AE in rats impacts the relationship between PFC structure, function, connectivity, and behavioral flexibility (namely, rule switching ability).

4.3. Potential for sex differences following developmental alcohol exposure in rodents

While the current study only examined male rats, further examination of potential sex differences following developmental alcohol exposure must be explored, with subsequent studies not only including females, but also including sufficiently powered sample sizes to detect such differences (we have provided observed effect sizes and achieved power in the analyses for the current study). Previous studies examining prefrontal and hippocampal alterations following early postnatal alcohol exposure in rats have been inconsistent in finding sex-related differences. While some literature did not find significant sex differences in molecular or anatomical measures within PFC or HPC [15, 16], several behavioral studies have observed differential influence of alcohol between sexes [20, 60, 61]. Re damage has been observed in females [25], and permanent Re damage following early postnatal alcohol exposure in rats does not vary between sexes [26], resulting in our hypothesis that damage selective to the mPFC-Re-HPC circuit is a mechanism by which alcohol exposure during the brain growth spurt results in impairments in executive functioning. Although Re vulnerability does not appear to be sex-dependent [26], the inconsistency of sex differences in past FASD-related behavioral research indicates an imperative for investigating sex as a biological modulator of alcohol-related brain damage, especially when correlating mPFC-Re-HPC structure, function, and connectivity with behavioral outcomes (especially EF-related behaviors) throughout life.

5. Conclusions

The convergence between behavioral impairments observed in the current study using a rodent model FASD and human studies of individuals diagnosed with FASD suggests that animal models show a high degree of validity for the human diagnostic condition. A comprehensive characterization of behavioral outcomes following different patterns of developmental AE is essential, as FASD is often misdiagnosed (as another disorder) or missed in diagnosis altogether [4]. The specific behavioral alterations observed in our rat model of binge alcohol exposure during human third-trimester suggest that damage to brain structures supporting executive functioning are most likely to result from aberrant connectivity of the mPFC-Re-HPC circuit, as supported by Re-specific damage to midline thalamus in the same rodent model [25]. These convergent findings necessitate further examination into a circuit-level analysis of brain dysfunction following developmental alcohol exposure.

Highlights.

• Developmental alcohol exposure disrupts memory for object placement

• Developmental alcohol exposure spares spontaneous alternation, object recognition

• Discrimination learning and reversal learning are not influenced by developmental alcohol

• Developmental alcohol exposure impairs rule switching

• Rule switching impairment implicates deficits in thalamic maintenance of mPFC learning

Abbreviations

AE

alcohol-exposed group (manipulated group)

BAC

blood alcohol concentration

FASD

Fetal Alcohol Spectrum Disorders

HPC

hippocampus

mPFC

medial prefrontal cortex

NOR

novel object recognition

OIP

object-in-place task

PD

postnatal day

Re

thalamic nucleus reuniens

SI

sham intubated group (procedural control group)