A new study from the University of Chicago and Johns Hopkins University maps gene-by-environment interactions across many cell types in response to common exposures like alcohol, nicotine, and caffeine. In the future, researchers hope to use such interactions to predict how environmental factors will affect an individual person.
"People can be exposed to the same environment. Some develop a disease, whereas other people in that environment do not," said Yoav Gilad, PhD, Professor of Medicine and Human Genetics at UChicago and co-senior author of the study, published in Cell Genomics.
Such environments can include any non-genetic factor, like chemical exposures and lifestyle choices.
One of the biggest challenges in human genetics, is to understand why those environmental effects differ from person to person. That's what we call gene-by-environment interactions."
Yoav Gilad, PhD, Professor of Medicine and Human Genetics, UChicago
The other 99% of your DNA
Gene-by-environment interactions are thought to underlie many of the differences in gene activity that contribute to disease risk. Any two unrelated people differ from each other at a few million locations in their DNA. RNA is made from protein-encoding genes and used as a template to make proteins. But only approximately 1% of the human genome codes for proteins. What about the rest of your DNA?
Many genetic variants linked to disease lie in regulatory regions. These regulatory regions act like a dimmer switch, controlling how much a gene is turned "on," without changing the nature of the protein itself. The amount of a gene's RNA in each cell tells you how much its expression is dialed up or down. Scientists want to understand how genetic variants affect the level of gene expression.
Filling in the matrix
The human body has hundreds of distinct cell types, and genes are switched on or off to help cells serve specific functions. As Gilad explained, "The challenge here is that those molecular interactions are going to be different depending on the cellular context: the cell type, developmental stage, and tissue."
To visualize the scale of the problem, he suggested picturing a large matrix. "On one axis are different cell contexts, and on the other axis millions of exposures, even combinations of exposures. We need data to be able to start filling that matrix and generating predictions," Gilad said.
Previous studies have characterized the impact of an environmental factor, like nicotine, in a particular cell type or tissue, with the goal of filling in one square in the matrix at a time. But that doesn't necessarily tell you what will happen in another cell type exposed to the same environment. That prompted Gilad's team to develop a method that characterizes gene-by-environment interactions in cells from many different people across many contexts at the same time.
Developing a cellular census
To study cells of different types and developmental states, the researchers used what they call "heterogeneous differentiating cultures." Adult donors provided samples through the International HapMap project. The Gilad lab reprogrammed these adult cells so that they could become any cell in the body, forming 3D clumps that represent diverse cell types in the early stages of development. They treated the resulting cultures with common environmental exposures, pooled cells from many donor samples, and traced each cell back to its donor.
"Until about 10 years ago, you couldn't separate those cells and study them individually," Gilad explained. "But the emergence of single-cell technology now allows you to characterize the RNA in each of those individual cells. Based on the RNA, you can tell which cell it was, identify the cell type, and then figure out how the cell responded to the treatment." By treating these cultures with alcohol, nicotine, or caffeine, the team tracked how each cell type responded across different genetic backgrounds, ultimately profiling over 1.4 million cells.
Exposing hidden effects on gene expression
In collaboration with researchers from Johns Hopkins, including co-senior author Alexis Battle, the Gilad lab searched for variants in DNA that are linked to gene expression levels. For each variant, they compared gene expression from different people, factoring in cell type. A variant that correlates with the level of a nearby gene's RNA is called an expression quantitative trait locus, or eQTL.
The standard, or baseline, eQTLs are genotypes that are associated with differences in gene expression between individuals, but are not sensitive to particular treatments. If the association changes with a treatment, it is called a response eQTL.
In a key finding, the researchers discovered that baseline eQTLs tend to cluster near the start of a gene, while response eQTLs are more likely to sit in enhancer regions, which boost a gene's activity.
"This is one of the first papers that showed the difference between the two in terms of their genomic features and locations," Gilad said. "The response eQTLs are in different genomic contexts, and drive the more dynamic nature of gene regulation, rather than just switching a gene on and off."
Finding response eQTLs
Response eQTLs show up at the same location as disease-linked variants much more often than baseline eQTLs. Gilad noted the importance of understanding the different types of eQTLs in explaining disease. "The context-specific associations are often much more relevant for diseases and clinically important phenotypes than the baseline associations," he said.
The researchers identified a response eQTL that impacts the gene LNX2 after treatment with alcohol, specifically in the type of cells that line the intestine. The eQTL is located in an enhancer, and it overlaps with a risk signal for inflammatory bowel disease (IBD). IBD causes chronic inflammation of the gastrointestinal tract, leading to damage of the intestinal lining. LNX2 is linked to a signaling pathway that helps build and maintain the gut lining, and future work could explore the molecular basis of its possible role in IBD.
While this does not mean that drinking alcohol will cause IBD, it points to how the effects of different genetic backgrounds can depend on both the exposure and the cell type. Gilad noted that "the reason that we want to be able to characterize this is because we want to be able to do more individualized prediction."
Predictions for the 101st patient
Making better risk predictions involves characterizing gene-by-environment interactions, including identifying where a treatment is working and any off-target effects.
"Even in treatments that we characterize in 100 individuals, what you really want to do is predict the response of the 101st," Gilad said. With training data that reflects both developmental and environmental contexts, AI models could provide relevant predictive insight in the future.
Predictive medicine could benefit people who are not represented in clinical trials. As an example, Gilad pointed out that there are very few treatments that have been tested on pregnant women. As a result, when a pregnant person is sick, "every clinician has to balance their intuition and anecdotal evidence" to weigh risk during treatment, "which is also going to be an individual risk based on genotype."
"A gene-by-environment interaction during development will not necessarily immediately translate to something meaningful in early childhood," Gilad said. "It could be that now there's some developmental quirk that underlies a higher risk of diabetes or heart disease down the road in your 40s or 50s." This study is a first step toward collecting the data that will back predictions on how these types of exposures ultimately impact individuals throughout their lives.
Source:
Journal reference:
Lin, W., et al. (2026). Disease-associated loci share properties with response eQTLs under common environmental exposures. Cell Genomics. DOI: 10.1016/j.xgen.2026.101383. https://www.sciencedirect.com/science/article/pii/S2666979X26002454?via%3Dihub