Is Vulnerability to Cocaine Written in Our Genes?

Cocaine use disorder is one of the most stubborn problems in addiction medicine. Roughly 1.4 million Americans meet the criteria for it, and after decades of research there is still not a single FDA-approved medication to treat it. We have known for a long time that genetics plays a role. Twin studies put the heritability of cocaine dependence as high as 70%, among the highest of any psychiatric condition. And yet, if you ask which genes are actually involved, the honest answer until recently has been that we barely know. Human genetic studies of cocaine use disorder are still relatively small, and only a handful of genes have ever crossed the genome-wide significance line.

Our lab just published a paper in Nature Communications, led by Montana Kay Lara and Lieselot Carrette carried out together with Abraham Palmer's group, that tries to close part of that gap. It took nine years, several labs, and a small army of people who refused to give up on it. It is, to our knowledge, the largest genetic study of cocaine self-administration ever done in rats.

The question we started with was simple.

Can we find genes that predict which individuals are most vulnerable to compulsive cocaine use? Answering that takes two things: a lot of animals, and animals that differ from one another the way people do. Most cocaine self-administration studies follow a few dozen rats. We phenotyped 836 Heterogeneous Stock (HS) rats, an outbred population bred from eight founder strains specifically to capture the kind of genetic diversity you see across human populations. Each rat had extended access to cocaine over several weeks, and we tracked the features that matter most: how much they escalated their intake, how hard they would work for a single dose under a progressive ratio schedule, and whether they kept taking the drug when each infusion carried a chance of a mild footshock. That last measure, continued use despite punishment, is one of the closest things we have to the defining feature of severe addiction in people.

First, the behaviors hang together.

The rats that escalated their intake the most also tended to work the hardest for the drug and were the most likely to keep using despite consequences. These traits were tightly correlated and clustered into a common pattern, which is reassuring, because it means addiction-like behavior in these animals behaves like a coherent construct rather than a grab-bag of unrelated quirks. With this many animals, we could resolve that pattern more finely than before, separating early acquisition, late acquisition, escalation, and a motivational and compulsive factor that grouped together breaking point, responding during timeout, and responding under footshock. That structure makes biological sense, and it lines up with the Addiction Index framework we and others have been developing for years.

Second, this is heritable, and that is good news.

If you are going to hunt for genes, it helps to first confirm that there are genes to find. There are. SNP-based heritability for these behaviors came in around 7 to 16% depending on the trait. That is modest, and well below what twin studies estimate, but that gap is expected: SNP heritability only captures the contribution of common genetic variants, not the full genetic picture. The reassuring part is that the most heritable trait of all was a composite measure of behavior during the escalation of intake, and seven of the ten most heritable traits came from the long-access phase. In other words, the genetic signal got stronger the deeper the animals went into addiction-like behavior, which is a nice validation of why we use the long-access model in the first place.

Then we went looking for the actual genes.

We found six genetic loci, spread across five chromosomes, each tied to a different aspect of addiction-like behavior. Two things stood out.

Some of the genes were old friends. We picked up Trak2, one of the very few genes that has been linked to cocaine use disorder in human genetic studies, and which helps traffic GABA receptors in the brain, a system we and others have repeatedly found to be central to drug self-administration. We also replicated several genes previously tied to alcohol and tobacco use in humans, including SLC10A7, PLCL1, and SATB2. Finding the same genes in rats and in people is rare, and it is exactly the kind of cross-species agreement that tells us the model is pointing at real, shared biology.

One locus was unusually clean. It sat over a small family of carboxylesterase genes, the enzymes that break cocaine down in the body, with a direct counterpart to the human enzyme CES1. It was associated with the time a rat left between infusions, a measure long thought to reflect the compulsive pacing of drug intake, and the effect was large enough to nudge an individual from a milder pattern of use toward a more severe one. This echoes human genetic studies, where enzymes that control how the body processes a drug keep turning up for alcohol and nicotine. It points to something we do not talk about enough: how fast you metabolize a drug may shape how vulnerable you are to it.

And because no genetic study is complete without a few surprises, we also flagged a set of brand-new candidate genes with effects on gene expression or splicing in addiction-relevant brain regions, including Rasd2, Gnas, Ctsz, Lsm6, Vsnl1, Zfp831, and Slc6a2, many of them tied to dopamine and norepinephrine signaling. These are leads we did not have before, and several of them are plausible targets for future treatments.

Why does this matter beyond the lab?

There is a practical thread running through the metabolism finding. For more than fifteen years, several groups have tried to develop engineered carboxylesterases as a way to treat cocaine overdose and addiction, the idea being that if you can speed up how fast cocaine is cleared, you can blunt its effects. That work has moved slowly, in part because the pharmaceutical industry has shown little appetite for it. What our data add is genetic evidence that variation in this very enzyme is tied to cocaine vulnerability in the first place. That gives the whole therapeutic direction more reason to exist.

It is also worth saying plainly, because it has become fashionable in some quarters to dismiss animal models as useless for understanding mental illness. They are not. This study is a clear example of rats and humans converging on the same genes and the same biology, and that convergence is how progress actually gets made.

From our lab's perspective, this is what we have been building toward.

We use HS rats precisely because their genetic diversity produces the spread you see in people: some rats escalate dramatically, some barely change, and everything in between. That variability is not noise to be averaged away. It is the signal. A single inbred strain, where every animal is genetically identical, would have erased exactly the differences this study was designed to find. Pairing that diversity with deep behavioral phenotyping is what let a metabolism gene, of all things, surface from something as simple as how rats paced their infusions.

What we still don't know is a lot.

We did not measure cocaine metabolism directly, so the link between these enzyme variants and how the drug is actually processed remains to be tested. A coding variant in a rat does not automatically correspond to a risk variant in a person. Our heritability estimates capture only common variants, our self-administration model uses intravenous delivery rather than the routes many people use, and we did not have the statistical power to look at how sex and genotype interact. These are real limitations. They are also a to-do list.

The bigger message is one I keep coming back to.

Addiction does not affect everyone equally, and the reasons why are written in our biology. Some of that biology is shared across species and across drugs, some of it is specific to the individual, and a growing amount of it is something we can now map gene by gene. Every time we find one of those genes, we get a little closer to treatments aimed at the people who need them most.

A word of thanks, and of honesty.

Nine years is a long time to spend on one study, and this one was hard in every way a project can be hard. I am grateful to the people in my lab who carried an enormous amount of the work, Marsida Kallupi, Giordano de Guglielmo, and Lieselot Carrette, to Abraham Palmer's group and Montana Kay Lara for the genetic analysis, and to Leah Solberg Woods at Wake Forest for breeding the rats that made all of this possible. I owe a particular thank you to our program officers at the NIH and to the study section reviewers who backed this project when it was far from a sure thing. Foundational work like this is slow, unglamorous, and easy to underfund, and it only happens when someone is willing to bet on it years before the payoff is visible. I am glad they did.

This work was supported by NIH grants P50DA037844, P30DA060810, U01DA051234, and U01DA043799. Full paper: Lara et al. (2026), Nature Communications. https://doi.org/10.1038/s41467-026-73694-w