A defining characteristic of alcoholism is compulsive drinking despite negative consequences. Thirty percent of Americans experience clinically defined alcohol use disorder (AUD) at some point in their lives. More than half recover, but that still leaves millions of people struggling with lifelong alcoholism in the U.S. alone. “Alcohol use disorders and excessive drinking kills more people than opiates,” says neuropharmacologist Kimberly Nixon, of the University of Texas at Austin. “People tend to forget that.”

Most people do not develop AUD, even if they drink heavily, or binge, but little is known about the factors that determine vulnerability to long-term compulsive drinking. A new study, led by pharmacologist Cody Siciliano of Vanderbilt University and neuroscientist Kay Tye of the Salk Institute for Biological Studies, describes a neural circuit that controls compulsive drinking in a mouse model. Activity in this circuit predicts weeks in advance which mice will go on to keep drinking despite negative consequences when given the chance. As well as identifying a brain-based “biomarker” for susceptibility to compulsive drinking, the study reveals a potential new target for the development of therapeutics for AUD, and possibly substance use disorders in general.

In the study, published Thursday in Science, the researchers first trained mice to associate a sound with delivery of sugar via a tube into their enclosure. Next, during a “prebinge” phase, the researchers replaced sugar with alcohol. They later gave the animals alcohol mixed with quinine (which has a bitter flavor, used as a punishment, or “aversive” consequence). During a “binge” phase, the mice had either no access to alcohol, or unlimited access to both water and alcohol for two or four hours a day. In a final “postbinge” phase the mice were again given alcohol, then alcohol-plus-quinine. The team divided the mice into groups based on their postbinge-phase alcohol consumption. “Low drinkers” drank little, whereas “high drinkers” drank a lot initially, but not once quinine was reintroduced. The third, “compulsive” group drank a lot, and was not put off by the added quinine. The researchers also used mild shocks as a punishment, and saw similar drinking patterns, showing these differences were not specific to quinine.

Tye and her colleagues suspected that two brain regions play a role in compulsive drinking. The medial prefrontal cortex (mPFC) is a higher brain region involved in behavioral control and other “executive” functions, such as judgment and decision-making (which are impaired in drug and alcohol use disorders). The dorsal periaqueductal gray (dPAG), is a brainstem region best known for its role in pain, but Tye and others had previously shown that neurons connecting the mPFC to the dPAG process aversive experiences. In the new study, the team used a high-tech imaging technique called calcium imaging to visualize the activity of hundreds of these “mPCF-dPAG” neurons at the start of the prebinge phase.

Although there were no differences in alcohol consumption at this stage between the compulsive mice and the other groups, they showed distinct neural activity. Activity in mPFC-dPAG neurons during the mice’s first licks of alcohol predicted which mice would develop compulsive drinking three weeks before the behavior actually emerged in a later phase of the study. “That’s what’s so exciting,” says Nixon, who was not involved in the study, but wrote an accompanying commentary in the same issue of Science. “This is some individual difference in those animals—and perhaps in people—that sets them up for severe symptoms down the road.”

The compulsive mice displayed a higher proportion of inhibitory, or brakelike, signals than excitatory ones, compared with the other mice. The researchers think this pattern of brain activity disrupts transmission of aversive signals in the mPFC-dPAG circuit, reducing sensitivity to punishment. To prove the circuit controls behavior in this way, they used a technique called optogenetics to make mPFC-dPAG neurons controllable using light. They showed that “turning off” the circuit increased compulsive drinking (the mice disregarded quinine more), whereas turning it on mimicked a punishment and reduced alcohol intake. “You get a lasting reduction in drinking, which is really exciting,” Tye says—“the idea that there’s some learning, or plastic change, that lasts beyond that specific day.”

The mPFC-dPAG brain circuit represents a new target for treatments. “This brainstem area has been very overlooked,” Nixon says. “It’s a circuit we hadn’t considered, that likely has novel pharmacological control systems we can look at.” Tye and colleagues have made a start. “We’re searching for druggable targets—small-molecule receptors on cell surfaces that would be unique to the circuit,” she says. “In the future, if we can induce synaptic plasticity in these circuits, we can potentially not only treat psychiatric disease states, but hopefully even cure them.” Identifying circuit-specific neural receptors will be important for avoiding side-effects.

Many fundamental questions remain, however. “There’s a lot more work to be done to understand the role of this circuit,” says neuroscientist Jeff Dalley of the University of Cambridge, who was not involved in the study. “What are the neurochemical players, what is the nature of these inhibitory and excitatory signals?” It is also not clear how relevant using quinine or shocks as a negative consequence are to the complex psychological phenomenon of human addiction, where consequences are typically far removed in time from the act. Tye claims there are analogues, though, such as barriers to consumption (the bar closing, or higher prices, for instance). “Those are immediate punishments a compulsive drinker wouldn’t think twice about,” she says.

The main caveat, however, is that scientists do not yet know how similar mice are to humans, in terms of signal encoding in the circuit, and the existence of a predictive biomarker. “That’s something we need to explore now,” Tye says. Nevertheless, “even the idea there could be a biomarker in the brain that predicts subsequent development of compulsive drinking is a big breakthrough we’re very excited about.”