1 Introduction
You are heading to dinner at a place near home. Halfway there, it starts raining, and you duck under a shelter. You consider whether you should go back home for an umbrella, wait a few minutes for the rain to stop, or just keep walking to dinner. What do you do? A crucial element of decision-making is flexibility, which implies the ability to adjust behavior in response to both environmental demands and personal factors (Luna, 2009; Diamond, 2013). However, this process requires complementary psychological functions like inhibitory control and working memory, among other components of executive function (Diamond, 2013).
Executive functions recruit subcortical and frontal, parietal and temporal cortical areas (Nowrangi et al., 2014), all of which are still developing during adolescence (Spear, 2013). Nevertheless, there has been a dominant interest in the role of the prefrontal cortex in executive function because of its involvement in “top-down control” (for review see Nowrangi et al., 2014; Dannenhoffer et al., 2018, in press). The prefrontal cortex matures later in development than other cortical regions (Luna, 2009; Spear, 2013), and this process might be altered by recreational drugs (Winters and Arria, 2011), including alcohol (see Spear, 2018; Crews et al., 2019 for review). Since executive function depends on adequate brain development, and flexible decision-making requires proper executive function, any potential hazard to brain development is detrimental to these processes.
Interest in the consequences of underage binge alcohol drinking has increased recently accompanied by many studies in humans and animals models (Dannenhoffer et al., 2021 in press). However, several gaps in knowledge remain, especially for humans, where a lack of longitudinal studies exploring brain circuits and cognitive processing affected by adolescent alcohol exposure is evident. Conversely, investigations in animal models have demonstrated the harmful and persistent effects that adolescent alcohol exposure can have on brain development and, consequently, on behavior. For example, adolescent intermittent ethanol (AIE) exposure can promote adult alcohol intake in male and female rodents (Gass et al., 2014; Amodeo et al., 2017), increase social avoidance (Dannenhoffer et al., 2018), perpetuate adolescent-typical behavior in adulthood (Spear and Swartzwelder, 2014), and impair performance in tasks that require inhibitory control, working memory and flexible decision-making (Schulteis et al., 2008; Salling et al., 2018; Crews et al., 2019; Macht et al., 2020). Moreover, AIE exposure increases conditioned approach to reward-associated cues (McClory and Spear, 2014; Kruse et al., 2017) and impairs flexible responses when experimental conditions change (Coleman et al., 2014; Gass et al., 2014; Fernandez and Savage, 2017; Galaj et al., 2019).
Consistent with those findings, we recently reported that AIE exposure shifted conditioned approach away from goal-tracking and toward sign-tracking during Pavlovian conditioned approach (Madayag et al., 2017). While the goal-tracking response generally adapts to changing conditions, such as reward omission, sign-tracking behavior is less flexible (Morrison et al., 2015; Ahrens et al., 2016; Stringfield et al., 2019). In additional studies, AIE exposure decreased reversal learning of an attentional set (Sey et al., 2019), indicating impaired inhibition of a previously learned response to a reward-predictive cue when the stimulus-response contingency changed. These behavioral tasks are governed by fronto-striatal circuits involved in action selection, and a separate study demonstrated that AIE exposure decreased functional connectivity of these frontostriatal circuits in rats (Broadwater et al., 2018). Together, these findings suggest a potential relationship between inflexible sign-tracking and inflexible attentional set strategy after AIE exposure, possibly mediated by changes in functional connectivity among prefrontal and striatal brain circuits involved in action selection. However, the fact that those results came from separate studies impedes the integration of these AIE effects. Here, we addressed this gap by testing the effects of AIE exposure on sensitivity to conditioned reinforcers (Pavlovian conditioned approach) and behavioral flexibility (attentional set-shifting) within the same animals and integrated them with resting-state functional connectivity MRI among frontal cortical and subcortical regions that regulate action selection. Then, using a dimensionality reduction approach, we identified a principal component of functional connections (functional subnetwork) to test potential mediational effects of functional connectivity on behavior. Thus, through this brain-wide MRI approach, we went beyond the typical analysis of discrete brain regions (i.e., nucleus accumbens, prefrontal cortex, hippocampus) or pairwise connections, which normally use electrochemical, electrophysiological, pharmacological and genetic approaches to understand the neural bases of behavior. Furthermore, the network analysis allows us to propose new circuits and connections that are involved in cognitive processing and decision-making and potentially affected by AIE exposure and to test whether they mediate the observed behavioral deficits.
Based on our previous studies, we hypothesized that AIE exposure promotes greater sensitivity to reward conditioning and inflexible action-selection in adulthood by altering shared underlying neural circuits. Specifically, we predicted that AIE exposure would increase bias to reward-associated cues and impair behavioral flexibility in adulthood when action outcomes change, and that those effects would be mediated by decreased functional connectivity after AIE exposure.
4 Discussion
Behavioral flexibility is a complex construct requiring several constituent functions that recruit multiple brain circuits. Thus, to investigate neural contributions to behavioral flexibility, we chose a network analysis approach over methods that focus on discrete brain regions. While prior studies reported that AIE exposure induces deficits in behavioral flexibility in adult rodents (Coleman et al., 2014; Gass et al., 2014; Madayag et al., 2017; Sey et al., 2019), the behavioral deficits could not be linked to changes in brain circuit function. In this study, we investigated AIE exposure effects on conditioned responding to reward-associated cues (Pavlovian conditioned approach), flexible decision-making (attentional set-shifting task), and resting-state functional connectivity MRI in the same animals to allow direct linkage of functional connectivity among brain regions involved in action selection to behavioral deficits. We did not observe AIE exposure effects on conditioned approach in this study, we did observe clear deficits in reversal learning and a marked reduction in functional connectivity among ROI. We also found a clear association between functional connectivity measures and behavioral performance. Moreover, we identified a functional connectivity subnetwork that mediated reversal learning deficits, demonstrating statistical causation between the degree of functional connectivity and behavioral flexibility. These complimentary analyses identified especially strong contributions of the NAc, thalamus, and OFC in the AIE-induced deficits in behavioral flexibility, a circuit associated with compulsive and inflexible behavior in substance use disorders (Volkow, 2000).
Behaviorally, we observed clear effects of AIE exposure on attentional set-shifting performance. It is worth noting that all observed deficits were identified after ∼55 days from the last alcohol administration and anteceded by PCA training. When considering various phases of the attentional set-shift task, phases implying new learning (e.g., initial acquisition) were not affected by AIE exposure, consistent with previous findings (Risher et al., 2013; Gass et al., 2014), showing that skills required to associate and discriminate stimuli are spared after AIE exposure. However, reacquisition was affected by AIE exposure, even though rats achieved the performance criterion the day before, suggesting deficits in consolidation. Before Reversal 1, AIE exposure impacted reacquisition differently by sex, with AIE-exposed females exhibiting more errors than control females, although the post-hoc comparisons did not reach significance. A potential limitation of this and other aspects of the attentional set task is that we used six consecutive correct choices as the performance criterion. While this criterion allowed us to directly compare our current results with previous reports (Birrell and Brown, 2000; Sey et al., 2019), it is possible that a more stringent criterion would reveal additional behavioral deficits.
We also observed AIE-induced deficits in both males and females during reacquisition before Reversal 2, indicating that although the new rule of Reversal 1 was learned, it was not as effectively consolidated among AIE-exposed animals compared to controls. As AIE-induced impairments in reacquisition performance were more robust before Reversal 2 (that assessed the reversed contingency) than before Reversal 1 (that assessed the original contingency), it is possible that learning of the original contingency was better consolidated in AIE-exposed rats than the reversed contingency. This interpretation is supported by the finding that AIE exposure did not alter reacquisition before the Extradimensional Set-Shift, which also assessed the original contingency. These results are only in partial agreement with our prior study (Sey et al., 2019), in which AIE exposure did not significantly impact any reacquisition phase. Reacquisition is not typically reported (Gass et al., 2014; Fernandez and Savage, 2017; Galaj et al., 2019), and we are not aware of studies evaluating the effect of AIE exposure on memory consolidation. However, AIE exposure alters adolescent hippocampal development (Broadwater et al., 2014; Vetreno and Crews, 2015), a critical region involved in memory consolidation through interaction with other cortical and subcortical regions (see McIntyre et al., 2011 for review).
The introduction of reversal phases revealed interesting differences among AIE-exposed and control groups. In Reversal 1, we observed no differences regarding total trials to criterion or errors, although prepotent responses were marginally different in that AIE-exposed subjects required more trials to adapt to the new rule. In Reversal 2, AIE-exposed rats made more active errors, and specifically more prepotent responses and subsequent regressive errors, while learning the new rule. Thus, AIE-exposed rats were slower to incorporate negative feedback and initiate a new strategy (Buisman-Pijlman et al., 2014) in both reversals, but this effect was stronger in Reversal 2. In contrast, in our previous study (Sey et al., 2019), AIE-induced deficits were observed in Reversal 1, while AIE-exposed rats performed better in Reversal 2. This pattern of behavior indicated that the original attentional set dominated behavioral choice, slowing the ability to learn an initial reversal but facilitating return to that contingency in Reversal 2. A major difference is that in the present study, rats had Pavlovian conditioning training prior to assessing attentional set learning, and previous behavioral training can impact subsequent task performance (Klintsova et al., 2002; Hannan, 2014; Fernandez and Savage, 2017). However, both studies identified increased prepotent responses in the AIE-exposed groups, consistent with perseveration on the previous response despite negative feedback. This particular deficit is suggestive of impaired inhibitory control (Luna, 2009; Adele; Diamond, 2013) and is consistent with other studies using AIE (Coleman et al., 2014; Fernandez and Savage, 2017; Galaj et al., 2019) and chronic intermittent ethanol (CIE) exposure in adult rodents (Badanich et al., 2011), and even in humans (for a recent review see López-Caneda et al., 2014). Moreover, the present study found that subsequent regressive errors were higher after AIE exposure. In general, regressive errors suggest a lack of ability to maintain a new association or keep the current rule “in mind” despite feedback.
Thus, our results suggest an AIE-induced deficit in working memory, a conclusion supported by other studies showing adolescent alcohol effects on that function in animals (Schulteis et al., 2008; Salling et al., 2018; Macht et al., 2020) and, less consistently, in humans (Salas-Gomez et al., 2016; Mahedy et al., 2018). Overall, these reversal data suggest that AIE exposure induces deficits in the fundamental processes of inhibitory control and working memory, resulting in deficits in flexible decision-making. Future studies can directly assess inhibitory control and working memory in animals with and without AIE exposure to confirm this interpretation.
Perseveration of an acquired attentional set illustrates how well-learned cues can retain control over action selection after AIE exposure. However, perseveration did not extend to the sensory domain of the relevant cue: when novel odor and tactile stimuli were used in the extradimensional set shift, all rats readily learned to pay attention to the tactile rather than the odor cue, consistent with previous studies (Badanich et al., 2011; Sey et al., 2019) using the digging task to assess alcohol effects on attentional sets. In contrast, studies using an operant-chamber version of a set-shift from a visual rule to a location rule reported AIE-induced deficits after chronic alcohol exposure (Hu et al., 2015) and AIE (Gass et al., 2014; Fernandez and Savage, 2017) [with one study only observing an AIE effect in males (Varlinskaya et al., 2020)]. However, in this latter task, the formerly relevant cues remain in the chamber, and the subject must inhibit responses to the visual cues and respond instead by location. Thus, those data can also be interpreted as an inability to inhibit a previously learned attentional set in AIE-exposed animals, as opposed to inability to shift the relevant stimulus from one domain to another.
While we replicated a sex difference in sign-tracking behavior in Pavlovian conditioned approach (Madayag et al., 2017), we did not observe an effect of AIE exposure on conditioned approach in the present study, in contrast to other reports in the literature (McClory and Spear, 2014; Kruse et al., 2017; Madayag et al., 2017). Large individual differences are often observed in conditioned approach due to considerable variability among animals in the propensity to attribute incentive salience to cues (Flagel et al., 2007; Robinson et al., 2014). In addition, there is large variability in the field around Pavlovian conditioning methodology, such as the timing of the cue and the length of training, and these factors may have contributed to our outcome. Similar issues have arisen in the literature reporting “attentional-bias” to appetitive cues in humans, a behavior that is likened to Pavlovian conditioning. Variations in the methods used to evaluate attentional bias, such as task psychometric properties, sample size, or cue sensory modality, among others, potentially explain inconsistent results among studies (Cisler et al., 2010; Field et al., 2010). While we did not replicate our previous findings, the majority of studies (McClory and Spear, 2014; Kruse et al., 2017; Madayag et al., 2017) to date have reported shifts toward sign-tracking after AIE exposure, suggesting enhanced sensitivity to reward-associated cues or incentive salience attribution. It’s worth noting that we did not test the persistence of conditioned approach behavior (e.g., via reward omission) or the development of incentive salience of the conditioned cue (e.g., via Pavlovian instrumental transfer). Thus, an empirical question is whether flexibility in these aspects of reward conditioning is altered by AIE exposure.
Action selection and behavioral flexibility in response to changing circumstances are largely mediated by cortico-striatal-thalamic circuits (McKim et al., 2016). Moreover, gaps in memory encoding or consolidation might be associated with reduced function of prefrontal and hippocampal areas (Carbia et al., 2018; Takehara-Nishiuchi, 2020), deficits in inhibitory control with fronto-striatal regions (Diamond, 2013; López-Caneda et al., 2014), and impairments in working memory and behavioral flexibility mainly with prefrontal regions (Diamond, 2013; Crews et al., 2019)—all brain regions that have been shown to be affected by AIE exposure (Crews et al., 2019). Thus, we hypothesized that behavioral deficits would be associated with altered functional connectivity among fronto-limbic regions in AIE-exposed animals. We found that AIE exposure reduced functional connectivity across multiple ROI pairs, including frontal, hippocampal and thalamic connections with the NAc. We previously observed similar AIE-induced reductions in functional connectivity across fronto-striatal regions of male rats (Broadwater et al., 2018), and the present study extended those findings to females. Indeed, the present findings suggest a robust disruption of NAc functional connectivity. Independent studies have shown the crucial role of this brain region in behavioral flexibility (Radke et al., 2019; Korn et al., 2021), and other studies have reported effects of AIE exposure on NAc physiology (Spoelder et al., 2015; Shnitko et al., 2016). Thus, a major contribution of the current study was to determine multiple specific connections with the NAc that were statistically linked to behavioral flexibility as well as affected by AIE exposure.
Linking the MRI data to behavioral assessments, the correlational analysis between pairwise connectivity strength and behavioral performance in Reversal 2 showed a negative association between functional connectivity in specific ROI pairs (including OFC, NAc, CPu, and Thal) and observed behavioral deficits. Interestingly, the correlation was stronger in water-exposed rats and reduced in AIE-exposed rats, suggesting that ethanol disrupted the association between functional connectivity and the behavior. Since we were underpowered to statistically validate this statement, additional studies including more animals can clarify this effect. An additional aspect derived from this analysis was the alleged key role that connections between OFC, NAc, and Thal play in the effects of AIE exposure on behavioral deficits, something also observed in the mediation analysis.
To address the interconnected nature of the ROIs, we used principal component analysis of functional connections to identify a principal component or “subnetwork” that might mediate observed behavioral deficits. The functional connections of the identified subnetwork (accounting for 66% of variance) include brain regions and projections observed in several studies to be involved in the cognitive processes of flexible decision-making (Jongen-Rêlo et al., 2003; Seamans et al., 2008; Yoon et al., 2008; Luna, 2009; Moore et al., 2013; Wang et al., 2015; Davies et al., 2017; Hardung et al., 2017; Esmaeili and Diamond, 2019). Interestingly, two connections that contributed most strongly to this subnetwork—NAc-Thal and Thal-OFC—form the striato-thalamo-orbitofrontal circuit that is hypothesized to be persistently impaired after repeated exposure to drugs of abuse and to contribute to the compulsive and inflexible behaviors associated with addiction (Volkow, 2000). The finding that adolescent alcohol exposure, in the absence of “addiction,” can affect similar circuits as those affected by chronic substance use supports an apparent vulnerability of prefrontal structures implicated in “top-down” subcortical regulation to drug exposure (Volkow, 2000; Spear, 2018; Crews et al., 2019).
Using this subnetwork as a mediator, we determined that AIE-associated decreases in functional connectivity mediated the increases in active errors during Reversal 2, when both contingencies had been experienced. While at least one prior study has correlated specific MRI measures to animal behavior (Li et al., 2018), to our knowledge, this is the first study to demonstrate that changes in functional connectivity statistically mediated behavioral performance in the same set of animals. Thus, it is validating that the different approaches to assess the impact of AIE exposure on brain functional connectivity yielded converging results highlighting connections between prefrontal, accumbal and thalamic regions. These findings also support future studies to explore individual pathways identified in the present data as relevant to cognitive flexibility (NAc-Thal and Thal-OFC) and its involvement in behavioral deficits after adolescent alcohol exposure.
There are many possible mechanisms that may underlie this altered functional connectivity following adolescent alcohol exposure, one of which is AIE-induced changes in cholinergic function. There is compelling evidence showing decreases in forebrain cholinergic phenotypic markers (Ehlers et al., 2011; Coleman et al., 2014; Fernandez and Savage, 2017; Galaj et al., 2019) and receptors (Coleman et al., 2014) after AIE exposure in rodents, supporting AIE-induced reductions in cholinergic neurotransmission that persist into adulthood and likely affect behavior (see Crews et al., 2019 for a review). Basal forebrain cholinergic neurons project extensively throughout the brain, including several regions in the functional connectivity subnetwork we identified: prefrontal cortex, hippocampus, and sensory cortex receive direct, modulatory cholinergic projections from the basal forebrain (Picciotto et al., 2012; Muñoz and Rudy, 2014). Documented cholinergic involvement in cognitive processes like learning, attention and memory [including working memory (Ballinger et al., 2016; Eckart et al., 2016)] implicates acetylcholine as a crucial player in cognitive performance. Thus, changes in cholinergic dynamics might explain behavioral and connectivity impairments in AIE-exposed animals; however, this hypothesis remains to be directly tested.
A strength of this study is large number of animals included, particularly for the behavioral measurements, which allowed for assessment of sex as a biological variable. Although no statistical differences were observed in functional connectivity among males and females, we found that females displayed a stronger deficit in reversal learning after AIE exposure and more sign-tracking behavior relative to males. While a sex difference in sign-tracking behavior is consistent with the literature (Pitchers et al., 2015; Sey et al., 2019; Stringfield et al., 2019), many studies assessing reversal learning after AIE exposure have only investigated cognitive deficits in males (Gass et al., 2014), were not sufficiently powered to address sex differences (Sey et al., 2019), or presented similar AIE effects between males and females (Galaj et al., 2019). However, understanding the underlying mechanisms that are sex-specific, such as immune reactivity and maturation of neural networks (Bilbo, 2018), may explain the domain, severity and persistence of cognitive deficits in females versus males.
It is important to note that animals were “passively” exposed to alcohol via experimenter-delivered gavage through a widely used protocol designed by the NADIA consortium (Crews et al., 2019) and tested in our lab (Madayag et al., 2017; Sey et al., 2019) and others (Vetreno and Crews, 2015; Dannenhoffer et al., 2018). This protocol ensures binge-like blood ethanol concentrations, modeling intermittent binge drinking that is often observed in human adolescents who drink. While beyond the aims of this study, it is possible that different results would be obtained if rats consumed the ethanol, as some behavioral and neurochemical differences have been observed between experimenter-delivered and self-administered alcohol (e.g., Weise-Kelly and Siegel, 2001; Orrù et al., 2016). A limitation of the present study is that animals were sedated during MRI scans rather than awake. While awake rat fMRI is possible, we used the most widely used light anesthesia protocol (Fukuda et al., 2013) to avoid the potential confounds of habituation and restraint after our behavioral approaches. Importantly, studies have shown that major brain connectivity changes are preserved across anesthesia states (Ma et al., 2017). Finally, while this study documented AIE-induced changes in behavioral flexibility and network connectivity several weeks after ethanol exposure, future studies can determine whether these effects of ethanol are specific to adolescent exposure and how they change across the lifespan.
In summary, we identified a functional connectivity subnetwork based on selected ROIs that was highly predictive of behavioral performance and mediated the effect of AIE exposure on reversal learning. This also suggests an effect of adolescent alcohol exposure on the striato-thalamo-orbitofrontal circuit, which has been related to compulsive behavior and risk for developing substance use disorders (Volkow, 2000). Here, we provided support for the hypothesis that decreased functional connectivity after AIE exposure mediates deficits in cognitive processing that impair flexible decision-making. Moreover, by using a network analysis approach, we provide evidence about AIE effects on the functional system that underlies complex processes involved in flexible decision-making, adding a fundamental piece for the understanding of binge-drinking effects on brain development and function.