According to the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), drug-induced liver injury (DILI) is the leading cause of acute liver failure in the U.S.

It is also a leading cause of drug failure in clinical trials.

Now, researchers have used liver organoids to develop a polygenic risk score that could predict the risk of liver toxicity for multiple different drugs, regardless of the underlying mechanism.

“Multiple drugs’ risk injury can be projected in our single model, despite the fact that the liver injury may be caused by different molecular mechanisms for each drug,” Takanori Takebe told BioWorld. Takebe is an assistant professor of pediatrics at the University of Cincinnati, as well as the director for commercial innovation and the university’s Center for Stem Cell and Organoid Research and Medicine. He is also a professor at Tokyo Medical and Dental University.

He and his colleagues reported their method in the Sept. 7, 2020, online issue of Nature Medicine.

Liver metabolism differs enough among species that only about 70% of drugs that end up with liver toxicity issues show any kind of warning signs in animal studies. As a result, liver injury has sunk a number of drugs in late-stage clinical development, or even after approval.

There are a number of gene variants that have been linked to an increased risk of drug toxicity. The cytochrome P450 enzymes, which play major roles in drug metabolism, were first recognized as related to risk in the 1980s.

More recently, genome-wide association studies (GWAS) have identified a number of variants that are more weakly linked to DILI.

Like in most conditions, though, GWAS has failed to deliver smoking guns, instead identifying multiple weak contributors to risk.

Such multifactorial risk is much harder to use in patient stratification. But in other disorders, it has been possible to combine those multiple weak variants into “polygenic risk scores” that can quantify overall risk.

“The polygenic risk score-based approach is…. emerging for different diseases,” Takebe said. But to date, it has not been integrated with in vitro model systems. In doing so, Takebe and his colleagues have developed an approach that could find applications both in drug development and clinically.

The team first re-analyzed multiple previous studies to arrive at a risk score that took more than 20,000 gene variants into account.

They then tested their model’s ability to predict what would happen in real, albeit in vitro, tissue.

When they exposed liver organoids to a dozen different drugs, the model was able to predict organoid susceptibility to DILI.

The organoids do not contain immune cells, and other studies have identified immune variants as predictive for DILI.

“Once the liver is damaged, adaptive immunity is targeting liver cell inflammatory processes,” Takebe said, and those will not be directly captured in hepatocyte-based risk scores. But “the trigger of liver damage can be modeled.”

The investigators also showed that the model was able to predict a high DILI risk in patients. The team recruited participants in a clinical trial for fasiglifam (TAK-875, Takeda Pharmaceutical Co. Ltd.), an experimental diabetes drug that failed in phase III due to DILI. Preclinical models had not predicted the liver toxicity risk. But when Takebe and his team applied their polygenic risk score to trial participants, their model was able to retrospectively identify those who had had an adverse reaction.

Several of the paper’s co-authors are Takeda employees, and the company provided some of the funding for the work. Takebe said that his team has “handed over the screening process to their scientists” for validation.

“If that is proven to be effective, they want to take this system as a routine process to predict DILI,” he said. “That is the obvious next step.”

But the model’s utility is “not just preclinical,” he stressed.

Patients and doctors “could potentially use our polygenic risk score as a diagnostic tool” to help make decisions for which drugs are likely safe, or not, for a patient.

If a patient has been genotyped by a commercial service such as 23andme, he said, “you can determine your own risk score, based on our study.”

No Comments