Alcohol use disorder (AUD) constitutes a global problem and is ranked among the top substance abuse problems in the United States, with over 70% of adults that struggle with substance use disorder estimated to abuse alcohol (Substance Abuse and Mental Health Services Administration (SAMHSA), 2020). AUD impacts global brain functional networks including the default mode, executive, attentional, salience and reward networks (Loeber et al., 2009; Park et al., 2010; Chanraud et al., 2011, 2013; Camchong et al., 2013; Sullivan et al., 2013) but the neurocircuitry underlying vulnerability and resilience to AUD is not clearly understood, making it difficult to establish viable, targeted treatment options.

This lack of clarity is due, in part, to the difficulty in capturing an alcohol-naïve baseline in human subjects. Clinical studies are often conducted with long-term drinkers at different drinking phases after changes in brain networks have already manifested. It is clear that AUD is characterized in part by dysfunctional information processing (Sullivan et al., 2000) that occurs in part through altered brain activity during both resting state (RS) and task performance in alcoholics (Rangaswamy and Porjesz, 2008) as compared to other neurological conditions (Georgopoulos et al., 2007). Functional brain networks including the default mode, salience, and executive networks are known to be sensitive to chronic alcohol use (Sullivan and Pfefferbaum, 2005; Sullivan et al., 2013; Weiland et al., 2014; Fede et al., 2019) however, the temporal nature and anatomic directionality of changes that occur remains unclear.

Studies such as IMAGEN have attempted to address this issue by evaluating adolescents prior to initiation of substance use (Maričić et al., 2020). Ivanov et al. (2020) used the IMAGEN study data to demonstrate that expected reductions in impulsivity were associated with lower levels of new onset drinking in an adolescent sample. They also found that blunted medial orbitofrontal activity in response to reward was associated with increased new onset and use. Nees et al. (2012) used the IMAGEN dataset to demonstrate that personality factors (novelty seeking, impulsivity, extraversion, and sensation seeking) were more strongly related to the initiation of drinking behaviors than reward related brain function, but that reward related brain function (particularly ventral striatum) was more strongly related to the development of problematic alcohol use. Heinrich et al. (2016) extended these findings in the IMAGEN study by including genetic factors. Results replicated the strong influence of personality factors on early initiation of drinking behaviors, while genetic factors appeared to become equally important when predicting future alcohol misuse behaviors. Reward related brain function did not contribute significant variance to the model. Similarly, Harper et al. (2019) used data from the Minnesota Twin Family Study Enrichment Sample to demonstrate that P3 amplitude mid-frontal theta power during an oddball task were predictors of new onset alcohol behaviors 3 years later in a sample of 14-year-old adolescents. While these studies offer valuable insight into the factors associated with initiation of use and onset of problematic use, they do not fully alleviate the confounding influence of environmental variables.

Non-human primate (NHP) models are valuable tools to help address the limitations of studies involving human participants (Grant et al., 2008). The model applied in the current study has been used to demonstrate that daily drinking for 15 months causes functional and genomic changes across the brain when contrasted against the alcohol naïve brain (Budygin et al., 2003; Floyd et al., 2004; Alexander et al., 2006; Carden et al., 2006; Anderson et al., 2007; Acosta et al., 2010; Cuzon Carlson et al., 2011; Mohr et al., 2013), as well as reorganization of brain networks measured by fMRI (Telesford et al., 2015) and significantly altered signal power of multiple bandwidths across the brain using magnetoencephalography (MEG; Rowland et al., 2017a). Using the same NHP model, alcohol naïve predictors of future drinking have also been identified. These include low cognitive flexibility (Shnitko et al., 2019), early drinking phenotypes (i.e., gulping vs. sipping; Grant et al., 2008; Baker et al., 2017), age, latency to begin drinking, and the number of “bouts” of drinking (Helms et al., 2014; Baker et al., 2017). No studies have examined alcohol naïve aspects of brain function as predictors of future drinking levels using this model.

The objective of the current study was a longitudinal examination of the trajectory of these changes at an earlier time point (6 months of chronic heavy alcohol intake) than previously examined (15 months chronic heavy alcohol intake) to identify at what point they begin to manifest and if baseline, alcohol-naïve indicators of future drinking can be identified.

Daily average g/kg EtOH consumption increased significantly once given free access [mean (SD); Free-Access = 4.7 (1.0), p < 0.01, Cohen’s d = 5.23], as shown in Table 1. Network metrics at Baseline and Free-Access (after 180 days unrestricted access) are shown in Table 2.

The current study demonstrates that characteristics of the Rich Club of alcohol-naïve brain networks are related to future drinking behaviors. In addition, following extended chronic drinking the connection strength in the network was altered; however, no effect on network metrics was observed. Finally, while hubs of the network were not observed to change significantly over time, membership in the Rich Club was significantly altered by chronic heavy drinking.

Aspects of alcohol-naïve resting-state functional brain networks were demonstrated to predict future drinking levels. Higher minimum degree of the Rich Club at Baseline (alcohol-naïve) was strongly correlated with decreased alcohol consumption during the future free access period. The Rich Club is a community of nodes within the network that have high degree and interconnectedness, serving as the “spine” of the network (Colizza et al., 2006). These nodes represent hubs within the network and communication among these nodes can often serve as “shortcuts” within the network, increasing efficiency of communication across otherwise distantly connected nodes. As the minimum degree of the Rich Club decreases, the centrality of Rich Club nodes also decreases, meaning fewer aspects of communication are routed through these nodes. Essentially, as the minimum degree decreases, the Rich Club becomes less differentiated from the other nodes in the network. This result suggests that the level of Rich Club differentiation may be predictive of future drinking levels, even when measured prior to alcohol exposure.

Alterations to the Rich Club subnetwork are likely to have broad and sweeping effects on brain communication and information processing (van den Heuvel and Sporns, 2013). Differences in Rich Club characteristics have been observed in many neurodevelopmental disorders, including schizophrenia (Collin et al., 2017), bipolar disorder (Wang et al., 2019), and autism (Hong et al., 2019). As such, the broad differences in Rich Club characteristics observed in this study are unlikely to serve as a direct “neurophenotype” of AUD without further refinement and empirical study. However, these results identify that differences in network topology are important to understanding individuals who might be at risk for future heavy drinking or AUD. Further, these results are consistent with previous work using the same NHP model indicating that premorbid behaviors and those occurring early in the drinking history may be predictive of future consumption levels. These factors include low cognitive flexibility (Shnitko et al., 2019), early drinking phenotypes (i.e., gulping vs. sipping; Grant et al., 2008; Baker et al., 2017), age, latency to begin drinking, and the number of “bouts” of drinking (Helms et al., 2014; Baker et al., 2017). The current results are the first to provide a brain-based factor indicative of future drinking in this model, suggesting that alcohol-naïve differences in Rich Club characteristics of functional brain networks also predict future drinking in this model.

These results extend recent findings in human participants demonstrating that white matter brain networks of individuals with AUD displayed lower Rich Club characteristics compared to their non-abusing siblings, who displayed lower levels compared to control participants (Zorlu et al., 2019). These results are also consistent with recent findings using the same NHP model indicating that chronic heavy drinking inhibits white matter growth (i.e., reduced connectivity) during late adolescence (Shnitko et al., 2019). While Zorlu et al. (2019) suggest potential premorbid differences in white matter network structure may be a marker of risk or susceptibility to AUD, the results of the current study provide direct empirical support for this hypothesis, showing that Rich Club characteristics of premorbid functional brain networks are directly related to future drinking levels. These results are also consistent with those of several longitudinal projects conducted in humans including the IMAGEN and NCANDA studies (Brumback et al., 2016; Maričić et al., 2020; Silveira et al., 2020). However, the current results examine functional RS brain networks and not reward based processing or structural aspects. The current study also did not acquire genetic or behavioral data as was acquired by IMAGEN, limiting further comparisons. Findings from the current study as well as these previous studies demonstrate that brain function and structure measured prior to exposure to alcohol may be able to identify individuals at risk for future heavy alcohol consumption.

Alcohol-induced changes in functional brain networks were observed following a period of exposure and free access to alcohol. The mean connection frequency increased following chronic heavy drinking; however, changes in network metrics were not observed. This suggests the general topology of networks were not altered (e.g., path lengths, clustering, etc.) Significant changes were seen in the membership of the Rich Club. There was much less consistency in Rich Club membership following chronic heavy drinking, suggesting the networks were being altered in an inconsistent manner across animals. It should be noted that the quantity of alcohol consumed during the free access period was fully determined by each animal. These results support the potential for a dose-dependent relationship between patterns of alcohol consumption and the effect on functional brain networks (Correas et al., 2015, 2016; López-Caneda et al., 2017; Perez-Ramirez et al., 2017).

Limitations of the current pilot study include the small sample size, which limits the complexity and sensitivity of analyses that can be conducted. Neuroimaging was conducted under anesthesia, which has known effects on brain function (Boveroux et al., 2010; Xie et al., 2013; Guldenmund et al., 2016). Possible interactions between the anesthetic and alcohol could have occurred, if not directly, then through the indirect development of tolerance. However, anesthesia was maintained at consistent levels and physiological indicators of arousal were monitored continuously, suggesting that levels of sedation were consistent across scans and animals. Additionally, after baseline data was acquired in the alcohol-naïve state, all animals entered the alcohol self-administration paradigm. We are mindful that this within-animal comparison across time limits insight into potential non-alcohol related changes over time or changes as a result of operant manipulations. However, test-retest scan sessions in monkeys 1 year apart and in humans 6 years apart (Stapleton-Kotloski et al., 2014) have established that networks remain stable over time.

Neuroimaging under conscious conditions will be required to completely understand the effects of alcohol on brain function using this model. The interval between ethanol access and MEG scans raises the possibility that some animals may have been experiencing symptoms of withdrawal (Winger and Woods, 1973). However, signs of withdrawal were not observed during the same time periods on non-imaging days (Pieper and Skeen, 1977; Cuzon Carlson et al., 2011). Also, propofol was used as the anesthetic agent, helping to ensure animals were not experiencing withdrawal symptoms during scans (Shin et al., 2018). Finally, animals who ceased alcohol consumption prior to scans did so voluntarily and in a time frame consistent with non-imaging days and were not forcibly fasted.

The current study identified a relationship between functional brain networks in the alcohol-naïve state and future alcohol consumption, consistent with other work using this model demonstrating early behavioral markers of future drinking. This is the first brain-based predictor of future alcohol consumption identified for this model. Additionally, significant alteration in the Rich Club of the network was observed following 180 days of chronic heavy consumption of alcohol, an earlier time point than examined in previous works. Future work will be invaluable in clarifying the changes, and specifying the timing of those changes, that infer risk specific to AUD in humans.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Wake Forest School of Medicine ACUC.

JR, JS-K, AD, DG, and JD contributed to the study design, JR, JS-K, AD, GA, PE, and JD contributed to data acquisition. JR, JS-K, DG, GA, and JD contributed to data analysis and interpretation. JR, JS-K, GA, AD, PE, DG, and JD contributed to manuscript preparation. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.