What Can Neuroimaging of Human Alcohol Drinkers Tell Us About Compulsive Alcohol Use?

A defining characteristic of alcohol use disorder is the continued use of alcohol despite negative life consequences, such as harmful social, financial, or health effects. This loss of control despite adverse consequences is defined as “compulsive” use and represents a major challenge for the treatment of alcoholism. For decades, researchers have been developing preclinical rodent models to simulate compulsive alcohol and drug use. These studies have used punishing stimuli—most commonly electric footshock (1) or conditioned taste aversion through the administration of quinine (2)—to test whether, and to what extent, animals are willing to undergo punishment to obtain a drug reward, a critical component of compulsive use. Generally, these studies have shown that the majority of nonaddicted animals reduce their drug intake when a punishment contingency is introduced, but some subgroups of animals do not respond to punishment (3). These “aversion-resistant” animals also show other phenotypes of alcoholism, such as escalation of use and a preference for higher concentrations of alcohol [see Hopf and Lesscher (4) for review]. More recent studies of aversion-resistant rats have identified neuroadaptations in brain regions such as the amygdala, neocortex, and striatum, identifying key biological pathways for compulsive drug use (4).

Compulsive alcohol use has not been modeled in human laboratory paradigms. In the current issue of Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, Grodin et al. (5) attempt to model compulsive alcohol seeking in human heavy drinkers (HDs) (defined as >20 drinks per week for males and >15 drinks per week in females), while these individuals underwent functional magnetic resonance imaging scans. In their paradigm, a modified version of the reward-based monetary incentive delay task was used, in which participants were presented with a series of trials with three reward cue symbols: a food reward, an alcohol reward, or neutral (no reward). To obtain these rewards, which were delivered immediately after the scan, participants could press a button as quickly as possible to hit a target. But there was a catch: each of these reward cues was paired with three potential threat levels of electric shocks to the wrist—green indicating safe, yellow indicating a low probability of shock, or red indicating a high probability of shock. If the participant chose to press the button to obtain the food or alcohol reward, he or she risked an immediate wrist shock (e.g., a negative consequence). A comparison group of light drinkers (LDs) also underwent all procedures. The authors’ main questions were as follows: would individuals be willing to endure electric shocks during the scan to obtain alcohol after the scan (e.g., would they seek alcohol despite a potential negative consequence)? Would the number of button presses differ between HDs and LDs? And what brain regions would show activation when individuals made the decision to respond for the alcohol in view of this negative consequence?

Surprisingly, both groups were willing to press for alcohol, even under high threat of electric shock, though HDs made significantly more button presses than LDs. HDs pressed for nearly 75% of high-threat alcohol trials, while LDs pressed for nearly 25% of high-threat alcohol trials. There was no difference between groups in the number of button presses during safe trials, or those in which there was no threat of shock. Interestingly, HDs also made more button presses than LDs to receive food or neutral rewards, indicating that perhaps HDs were less sensitive to punishment in general. This finding may have important implications. A recent Science paper investigating the mechanisms of cocaine addiction showed that in cocaine-addicted individuals, reward training improved response rates, but avoidance/punishment training (using electric shocks) had no effect on behavior (6). Together, these studies highlight that insensitivity to punishment may be an important characteristic of addiction, and on a policy level, these findings call into question the effectiveness of putative approaches in deterring compulsive drug use.

In Grodin et al. (5), neural activity during the presentation of high-threat alcohol trials partly mirrored the circuitry of compulsive use that has been identified in rodents. During high-threat alcohol cues, HDs showed greater activation than LDs in the striatum, medial prefrontal cortex (mPFC), and occipital brain regions. These regions were also more active in HDs during the threatening alcohol condition compared with the threatening food or neutral conditions, demonstrating some specificity to alcohol cues. Furthermore, a connectivity analysis showed that in HDs, greater connectivity between the anterior insula and the nucleus accumbens correlated with greater self-reported compulsivity. Activation reported in this study replicates animal work that has shown a proposed “compulsive network” in alcohol and cocaine addiction, comprising the mPFC, basolateral and central amygdala, nucleus accumbens, and dorsolateral striatum (4). Prefrontal cortical areas in particular have been thought to mediate compulsive behavior because the PFC is often involved in conflict (e.g., weighing the drug reward against the potential negative consequences) (7). The PFC is also important in craving and relapse (8). The striatum, another region showing differences between HDs and LDs during high-threat alcohol trials, is involved with reward-seeking or habitual drug use (9). In a clever preclinical experiment, Seif et al. (10) used optogenetics to manipulate the striatum during quinine and footshock paradigms, and inhibition of mPFC to nucleus accumbens inputs significantly reduced footshock-resistant responding for alcohol, providing clear evidence that cortical regions are necessary for aversion-resistant (i.e., compulsive) alcohol intake. It should be noted that many regions in the compulsive network overlap with those proposed for motivation/action networks, including the mPFC, amygdala, nucleus accumbens, and dorsal striatum. The activation reported by Grodin et al. (5) was reported in most of these regions as well, suggesting that the shift from impulsive to compulsive drug use likely involves much of the same brain circuitry.

The Grodin et al. (5) experiment had several limitations, most prominently that the neuroimaging paradigm likely did not isolate compulsive drinking. The HDs were likely drinking for a combination of reasons, including motivational, habitual, and compulsive reasons, and these likely varied across participants. Though many of the HDs had alcohol dependence, they were not treatment seeking, and therefore many of the negative life consequences associated with alcoholism may not (yet) have occurred for them. This experiment is also not a perfect model of human compulsive alcohol seeking. In compulsive drinking, the substance is used first, and negative consequences come later, while in Grodin et al.’s task, heavy drinkers were willing to endure a shock first and obtain the alcohol later. This sequence does, however, mirror preclinical work using quinine, where the bad taste of quinine is immediate yet the rewarding effects of alcohol intoxication are delayed. Second, and perhaps more fundamentally, real-life negative consequences of compulsive alcohol use are often delayed and are accumulations of many drinking episodes, whereas in the experiment the electric shock is an immediate and momentary punishment. Finally, the lack of sensitivity to adverse outcomes is a critical component of compulsive alcohol use, yet compulsivity also requires loss of control over substance use, which is not directly modeled in these experiments. Even considering these limitations, the results of this experiment suggest that the threat of adversity in the form of an electric shock may, in a small way, capture the internal conflict that individuals with alcoholism experience when contemplating a drinking episode.

Translational studies are critically important in the alcohol field, where the majority of what we know about the neurobiology of addiction comes from preclinical studies [e.g., Koob and Volkow (7)]. Preclinical studies have greatly enhanced our understanding of the shift from impulsive to compulsive use and have identified key brain regions and neurotransmitter systems that are important in addictive behavior. Despite this expansion in knowledge, alcohol addiction continues to be difficult to treat in humans, with high relapse rates. Grodin et al.’s work represents an ambitious and creative attempt to model compulsive alcohol use in a human neuroimaging study. While important methodological considerations limit interpretation of these findings, the data indicate that compulsive alcohol use in humans may involve similar brain circuitry as that in animals and may present targets in the treatment of alcohol use disorder. Understanding these brain systems in both healthy drinkers and in those with alcohol use disorder will increase our knowledge of the neural and behavioral components of the pathology of addiction.