A collaboration between scientists at The Jackson Laboratory (JAX), the Broad Institute of MIT and Harvard, and Yale University has led to the development of a new artificial intelligence-based method that designs DNA switches capable of controlling gene expression in specific cell types. This approach could improve the way researchers control gene activity, with potential applications in both basic research and therapeutics.
These DNA switches, known as cis-regulatory elements (CREs), play a critical role in gene regulation. CREs are sequences of DNA that, while not part of the genes themselves, influence when and where genes are turned on or off. The new research, published in Nature, describes how AI has been employed to design synthetic CREs that function with remarkable specificity to targeted cell types, including brain, liver, and blood cells.
Designing Cell-Specific DNA Switches
Gene therapies and genetic editing technologies have advanced significantly, but precisely controlling gene expression in specific tissues remains a challenge. This difficulty arises in part because CREs do not follow straightforward rules, complicating the task of designing genetic elements that act only in desired cell types.
The grammar of CREs is not fully understood, and there are no clear rules for what each CRE does. “This limits our ability to design gene therapies that only affect certain cell types in the human body,” explained Rodrigo Castro, a computational scientist at JAX and co-first author of the study. Using deep learning, the team was able to train an AI model on hundreds of thousands of DNA sequences and then predict the activity of these sequences in different cell types. These predictions allowed the researchers to identify patterns in CRE sequences and understand how they influence gene expression.
The new AI tool, called CODA (Computational Optimization of DNA Activity), used these patterns to design thousands of synthetic CREs. CODA allows researchers to fine-tune the design of CREs with specific characteristics, such as activating a gene in liver cells while leaving the same gene silent in other tissues.
Bringing AI-Designed CREs to Life
The synthetic CREs were tested in blood, liver, and brain cells to measure their effectiveness. The researchers found that these AI-designed elements were even more specific to their target tissues than naturally occurring CREs. This cell-type specificity is necessary for therapies where genes need to be activated or repressed in precise locations to avoid unintended effects elsewhere in the body.
“What is special about these synthetically designed elements is that they show remarkable specificity to the target cell type they were designed for,” said Ryan Tewhey, associate professor at JAX and co-senior author of the study. “This creates the opportunity for us to turn the expression of a gene up or down in just one tissue without affecting the rest of the body.”
One surprising finding was the effectiveness of these synthetic CREs despite their significant divergence from natural sequences. “The synthetic CREs semantically diverged so far from natural elements that predictions for their effectiveness seemed implausible," said Sager Gosai, a postdoctoral fellow at the Broad Institute and co-first author of the paper. "We initially expected many of the sequences would misbehave inside living cells." However, the results were highly promising, with the designed sequences achieving more precise control over gene activity than anticipated.
Testing the System in Living Organisms
To further validate the effectiveness of the synthetic CREs, the team tested them in animal models, including zebrafish and mice. The synthetic elements activated genes in the intended tissues without affecting others. In one experiment, a CRE designed for liver cells in zebrafish was able to turn on a fluorescent protein specifically in the fish's liver, leaving other tissues unaffected.
This level of control over gene expression opens up exciting possibilities for using CREs in biomedical applications. By creating synthetic elements tailored to specific cell types, researchers may one day use these technologies to control gene activity in a highly targeted manner, enabling therapies that minimize off-target effects.
Future Prospects for Precision Gene Regulation
The development of CODA and the success of the AI-designed CREs represent a significant advancement in the field of gene regulation. “This technology paves the way toward writing new regulatory elements with predefined functions,” said Tewhey. “Such tools will be valuable for basic research but also could have significant biomedical implications where you could use these elements to control gene expression in very specific cell types for therapeutic purposes.”
This approach to engineering genetic switches offers a way to bypass the limitations imposed by naturally occurring CREs, whose evolutionarily constrained functions restrict their flexibility.
As Pardis Sabeti, a core institute member at the Broad Institute and co-senior author of the study, noted, “These AI tools have immense potential for designing genetic switches that precisely tune gene expression for novel applications, such as biomanufacturing and therapeutics, that lie outside the scope of evolutionary pressures.”
Publication Details
Gosai, S.J., Castro, R.I., Fuentes, N. et al. Machine-guided design of cell-type-targeting cis-regulatory elements. Nature (2024). https://doi.org/10.1038/s41586-024-08070-z