The first goal of the current study was to investigate the extent to which AUD/CUD symptomatology level was associated with RPE signaling. In line with our hypotheses, greater AUDIT scores were associated with reduced modulation by RPE within ventral putamen (at a lenient extent threshold) and attentional regions such as dlPFC, iFG, aIC, and iPL, as well as with poorer performance on the task (i.e., proportion of best non-novel stimuli chosen). The second goal of the current study was to investigate the extent to which AUD/CUD symptomatology level was associated with RPE signaling to novel stimuli. We also found that during novel stimulus exploration, greater AUDIT scores were associated with diminished RPE modulation within ventral putamen, caudate, dlPFC, iFG, aIC, iPL, and at a lenient extent threshold, ACC. Finally, we found that NP was positively associated with RPE modulation within attentional regions such as iPL, dmPFC, and STG in individuals with subclinical CUDIT scores (0–6), but not in individuals with clinically significant CUDIT scores.

We predicted that AUD symptomatology level would be negatively associated with RPE signaling within striatum. This was because of previous work indicating that chronic substance use leads to hypo-responsiveness to non-drug reward related cues within the striatum in non-learning (Aloi et al., 2020, 2019; Beck et al., 2009; Tapert et al., 2001; Wrase et al., 2007) and instrumental learning (Aloi et al., 2020) paradigms. In line with this hypothesis, we found that greater AUD symptomatology level was associated with reduced RPE modulation within ventral putamen (at a lenient extent threshold). Most previous work looking at reward sensitivity in individuals with AUD has used the MID task (Aloi et al., 2019; Beck et al., 2009; Nees et al., 2015; Wrase et al., 2007). This task examines response to reward outcomes where reinforcement contingencies are determined by the task structure and known to the participants. Little work has examined reward responsiveness in the context of instrumental learning and, even so, analyses have not adopted a neuro-computational approach (Aloi et al., 2020). The current data extend the previous literature by showing that greater AUD symptomatology level is associated with reduced RPE representation within striatum. On the basis of these data, AUD symptomatology level is negatively associated with computations signaling the unexpectedness of received (non-drug) reward. Thus, instrumental learning and choosing actions that engender non-drug rewards should be compromised.

The current data indicate similar effects with regard to attentional regions, such as dlPFC, iPL, aIC, and iFG. Dorsolateral PFC and iPL comprise a central executive network that is associated with working memory and decision-making (Katsuki and Constantinidis, 2012), while aIC is implicated in salience detection and learning from uncertain rewards (Preuschoff et al., 2008). However, caution is warranted regarding reverse inference as the current study is unable to delineate the precise roles of these brain regions with regard to attentional subsystems in the Novelty Task. Prior literature has shown that AUD symptomatology level is associated with alterations within dlPFC, iPL, aIC, and iFG during tasks designed to elicit attentional neuro-circuitries (Ahmadi et al., 2013; Aloi et al., 2018; Thayer et al., 2015). However, previous work from our group has shown that in reward paradigms, reduced striatal responsivity in adolescents with greater AUD symptomatology level is also associated with reduced attentional responsiveness (Aloi et al., 2020, 2019). Moreover, prior work in adults has shown that alterations in dlPFC-striatal connectivity during an instrumental learning task in individuals with AUD predicted poorer performance on the task (Park et al., 2011). We hypothesize that the current data represent an alteration in orchestrating an attentional response to RPE.

Reduced RPE signaling should manifest behaviorally as poorer performance in reinforcement-based decision-making (Schultz, 2016). In line with our neuroimaging findings that AUDIT scores were negatively associated with RPE sensitivity, AUDIT scores were also negatively associated with behavioral performance on the current task. This aligns with previous work showing that individuals with AUD show poor performance in reinforcement-based decision-making tasks (Bechara et al., 2001; Tomassini et al., 2012) – though see (Aloi et al., 2020). Importantly, understanding alterations in neuro-computational processes in reward learning associated with AUD has significant clinical implications (García-García et al., 2017; Schultz, 2016). Reduced RPE responsiveness to non-drug reinforcers will result in these being less reinforcing, thus reducing the individual’s propensity to engage in actions that might engender these reinforcers—particularly relative to those associated with drug reinforcers where signaling may be increased (García-García et al., 2017; Schultz, 2016). Contingency management techniques based on principles of reinforcement learning are effective in adults with SUDs (Petry and Simcic, 2002). However, more thorough understanding of the mechanisms underlying these interventions in adolescents is needed given the maturation of neurobiological systems underlying reinforcement learning during this time period (Galvan, 2010; Nussenbaum and Hartley, 2019). Potentially, if interventions do not address these reinforcement-related difficulties, it will be difficult for these individuals to shift behavioral/attentional foci away from drug-related actions and towards other reward-related actions.

As noted above, striatum shows an augmented response to novel reinforcements in addition to its role in responding to RPE, as striatal RPE representation is enhanced during novelty-driven exploration (Costa et al., 2019; Wittmann et al., 2008). This function of striatum was also disrupted as a function of AUD symptomatology level; AUDIT scores were negatively associated with extent of increase in ventral putamen and caudate RPE modulation when exploring novel stimuli. This finding could be considered surprising since prior work has indicated that individuals with increased sensitivity to novel stimuli are particularly at risk for developing substance use behaviors (Cinnamon Bidwell et al., 2015; Ray et al., 2009). However, it is important to note the distinction between neuro-developmental risk factors for initiating substance use as opposed to the neuro-developmental impacts of substance use. Increased reward responsiveness and increased novelty seeking are risk factors for the development of substance use (Cinnamon Bidwell et al., 2015; Heitzeg et al., 2015; Hulvershorn et al., 2015; Ray et al., 2009). Yet, the current data, and considerable previous work (Aloi et al., 2020, 2019; Nees et al., 2015), indicate that AUD symptomatology level is associated with decreased reward responsiveness and, in the current study, reduced RPE signaling to novel stimuli. In short, there appear to be marked differences, at least with respect to non-drug reward sensitivity/RPE representation, in the neurobiological risk factors for substance use relative to the neurobiological consequences of severe substance (alcohol) use.

In initial analyses with healthy participants, we observed that the augmentation of the RPE signal to novel stimuli was particularly marked in individuals with greater NP (Fig. S2). In the current study, this was seen individuals with sub-clinical CUDIT scores within iPL, dmPFC, and STG. However, in individuals with high CUD symptomatology levels, there was reduced differential RPE modulation to novel vs. non-novel stimuli as a function of NP. This suggests that increased CUD symptomatology level is associated with a degree of compromised responsiveness to novelty, at least within regions implicated in attentional responding. Notably, previous work has shown that individuals with greater CUD symptomatology levels show reduced ACC/dmPFC responsiveness and awareness to error feedback (Aloi et al., 2019; Hester et al., 2009), which may be associated with “amotivational syndrome”, often observed in CUD (Pacheco-Colon et al., 2018). These data may indicate particular toxic impacts of cannabis use.

Prior work has suggested that individuals with SUDs show reduced striatal responsiveness to non-drug cues because chronic substance use results in dopamine receptor downregulation and reduced dopaminergic neurotransmission within this structure (Koob and Volkow, 2016). It has been suggested that chronic alcohol use and chronic cannabis use reduce synaptic dopamine to non-alcohol and non-cannabis stimuli, respectively (Koob, 2013; Martinez et al., 2005; Van De Giessen et al., 2017). Notably, however, molecular data indicates individual associations of neuronal impact for long-term alcohol and cannabis use. While both acute alcohol use and acute cannabis use induce increases in synaptic dopamine in the ventral striatum, the increase following alcohol appears larger (Boileau et al., 2003; Bossong et al., 2009). Moreover, individuals with AUD show reduced striatal dopamine receptor availability (Martinez et al., 2005; Trifilieff and Martinez, 2014) while individuals with CUD do not (Stokes et al., 2012; Trifilieff and Martinez, 2014). We believe our individual associations of AUDIT versus CUDIT scores with RPE may reflect these molecular effects; we speculate that prolonged alcohol use exerts a greater impact on striatal and cortical dopamine receptors long-term, leading to broader alterations in RPE representation relative to prolonged cannabis use.

The current study has several limitations. First, urine/breathalyzer testing for alcohol or cannabis use was not conducted at the time of scanning. However, all but two participants with significant alcohol and/or cannabis use histories were residents of a highly supervised residential youth care facility, and subject to random drug testing as part of the treatment program for at least four weeks prior to scanning. Exclusion of these two participants elicited highly similar results. Second, the present study was cross-sectional, so the results reported might reflect the effects of AUD/CUD on the developing brain or pre-existing risk factors for AUD/CUD. However, the current results reflect individual associations of AUD/CUD symptomatology level and RPE signaling; it is unclear that there are pre-existing neural risk factors that place individuals specifically at risk for AUD rather than CUD (or vice-versa). Third, there was a great degree of psychiatric co-morbidity within the sample. It could be argued that the findings in the current study are reflective of psychiatric co-morbidities of AUD/CUD rather than AUD/CUD itself. Notably, our supplemental analyses showed that including ADHD, CD, MDD, or GAD diagnoses as covariates did not significantly alter the main results. Additionally, a number of participants were prescribed psychotropic medications for these conditions. While this use might have influenced our findings, it is important to note that including stimulant, antidepressant, or antipsychotic use as covariates did not significantly alter the main results (see Supplementary Tables S7–S9). Therefore, the current findings likely reflect level of AUD symptomatology rather than any psychiatric co-morbidity or psychotropic medication use. Relatedly, AUD and CUD were highly co-morbid in our sample; of the 40 adolescents with AUDIT ≥ 4, only 9 did not have CUDIT ≥ 6. Such findings are common in adolescents; epidemiological data indicate that co-morbid alcohol/cannabis use is common (Mason et al., 2013). This makes interpretation of group-based studies very complex unless a clinically atypical single drug use group is identified. However, our dimensional approach enables identification of individual associations related to AUD versus CUD symptomatology levels. Nevertheless, future longitudinal work in groups of single drug use individuals will be necessary to determine causal effects of alcohol use only versus cannabis use only. Similarly, AUDIT and CUDIT scores were correlated, and so inclusion of both AUDIT and CUDIT scores could confound one another due to multicollinearity. However, the VIFs of the AUDIT and CUDIT regressors in our model were 1.49 and 1.42, respectively, suggesting that this was not the case. It should also be noted AUDIT and CUDIT are general indices of AUD and CUD symptomatology levels, respectively. These measures index rough levels of quantity/frequency, levels of abuse, and levels of dependence of alcohol and cannabis over the past year. However, additional indices such as age of first use and overall length of substance use, were unavailable for this sample. Future work is necessary to delineate the association between these important substance use parameters and RPE signaling. Finally, it should be noted that the Novelty task utilized monetary reward. A subject of future investigation should be whether these relationships between AUD symptomatology level and RPE signaling are consistent with regard to primary rewards (e.g., food). Relatedly, future work should also investigate the relationship between SUDs and RPE signaling with regard to substance-related stimuli. In particular, this work should focus on the relationship between AUD and RPE signaling with regard to alcohol-related stimuli.

In summary, AUDIT scores were inversely related to RPE signaling within ventral putamen, iFG, dlPFC, and iPL. We also found that in individuals with greater AUDIT scores, RPE signaling within caudate, ventral putamen, dlPFC, iFG, iPL, and ACC was particularly reduced when exploring novel stimuli. Furthermore, our data suggest that greater AUDIT scores were associated with weaker task performance. Regarding CUD symptoms, we found a positive relationship between NP and RPE modulation within dmPFC and iPL only in individuals with sub-clinical CUDIT scores. These data replicate prior work (Aloi et al., 2020, 2019) indicating that AUD symptomatology level is associated with alterations in striatal/cortical systems when receiving rewards and extend this finding by establishing that (i) AUD symptomatology level is associated with altered RPE signaling and (ii) this alteration is particularly pronounced when exploring novel stimuli.