High-Throughput Morphological Profiling of Patient-Derived Human Liver Organoids for Hepatotoxicity Hazard Identification and Compound Prioritization

The liver is the body's chemical processing plant, filtering toxins from the blood and breaking down medications so they can be used or safely discarded. Because this organ works so hard, it is also the most common site of damage caused by new drugs. When a medicine harms the liver, it can lead to severe illness or force a promising treatment to be pulled from the market, a failure that costs years of research and leaves patients without cures. For decades, scientists have tried to predict which drugs will hurt the liver before they ever reach a human patient. They have relied on tests using animal cells or simple human cells grown in flat layers on plastic. However, these old models often miss the mark because they lack the complex, three-dimensional structure of a real liver and the specific genetic makeup of different people. As a result, many drugs that look safe in the lab turn out to be dangerous in the clinic, while others that might have been safe are discarded too early.

A team of researchers at the University of Michigan has developed a new way to test drug safety that bridges this gap. Instead of using flat cells or animals, they grew tiny, three-dimensional clusters of human liver tissue, known as organoids, from stem cells taken from real patients. These tiny tissues contain multiple types of liver cells working together, much like a miniature version of the organ itself. The researchers then exposed these living tissues to a wide variety of drugs and used advanced cameras to take thousands of detailed pictures of how the cells looked and behaved. By training a computer to recognize subtle changes in the shape and structure of the cells, the team created a system that can spot early signs of liver damage long before the cells actually die. This approach offers a more human-like way to screen medicines, potentially catching dangerous side effects earlier and more accurately than ever before.

To build this system, the scientists started by collecting blood samples from seven individuals: six patients who had previously suffered from drug-induced liver injury and one healthy person with no history of liver problems. They reprogrammed cells from these blood samples into stem cells, which can turn into any type of tissue in the body. These stem cells were then guided to grow into liver organoids. Because the organoids came from real people, they carried the unique genetic fingerprints of their donors. This meant that the researchers could see how different people's livers might react to the same drug, capturing the natural variation found in the human population. The team grew these organoids, broke them down into single cells, and placed them into tiny wells on a plate, where they were treated with 74 different drugs. Some of these drugs were known to cause liver damage, while others were known to be safe.

After three days of exposure, the researchers fixed the cells and stained them with special fluorescent dyes that highlighted specific parts of the cell, such as the nucleus, the energy-producing mitochondria, and fat droplets. They then used a high-powered microscope to capture detailed images of thousands of cells in each well. The key innovation was not just taking the pictures, but how they analyzed them. Instead of simply counting how many cells were alive or dead, they used a computer program to measure hundreds of tiny details about the cells' shapes and textures. The computer learned to recognize specific patterns that signaled stress or injury, such as changes in the shape of the nucleus or the accumulation of fat. This method allowed them to detect subtle warning signs that traditional tests, which only look for cell death, would have missed.

The results showed that this new method was highly effective at distinguishing between dangerous and safe drugs. When the computer analyzed the images from the organoids, it correctly identified the harmful drugs with a high degree of accuracy, far better than tests that only measured cell survival. The system was so sensitive that it could spot early signs of trouble even when the cells were still alive. Interestingly, the researchers found that using a combination of organoids from all seven donors gave the best results. While each donor's liver cells reacted slightly differently, averaging their responses created a more reliable safety net than relying on just one type of cell. This suggests that testing drugs on a diverse panel of human tissues is crucial for catching the wide range of ways people might react to a medicine.

The study also revealed that the liver's reaction to drugs is not always the same for everyone. Some drugs caused similar changes in all the organoids, while others triggered very different responses depending on the donor. For example, one drug caused a massive shift in how the cells looked in some donors but had almost no effect in others. This variability mirrors what happens in the real world, where some people get sick from a medication while others do not. By measuring exactly how much of a drug was needed to cause these changes, the researchers could rank the drugs by their potential to cause harm. They found that for many drugs, the amount needed to cause damage in the lab was close to the levels found in human blood, suggesting the test is relevant to real-world exposure.

One of the most significant findings was that the system could detect specific types of liver injury based on the visual patterns it saw. The computer grouped the drugs into clusters based on how they changed the cells. One group of drugs caused the cells to look like they were struggling with their energy production, while another group caused the cells to fill up with fat or change their membrane structure. This ability to see the "signature" of the damage helps scientists understand not just that a drug is toxic, but how it might be hurting the liver. In one notable case, a drug called sulindac, which had previously been missed by the system, was correctly identified as dangerous when the researchers added a small amount of an inflammatory signal to the test. This showed that the system could be tweaked to catch drugs that cause injury only under specific conditions, such as when the body is already inflamed.

Despite its success, the researchers were careful to note the limits of their work. The system is designed to flag potential hazards and rank drugs by risk, not to predict exactly who will get sick or when. It works best as an early warning tool to help scientists decide which drugs to move forward with and which to discard. The study also highlighted that some drugs known to cause rare, immune-related liver damage were still difficult to detect because the organoids lack the full immune system found in a human body. However, the ability to add inflammatory signals to the test suggests a path forward for improving the system further.

Ultimately, this research demonstrates that growing tiny, patient-derived liver tissues and watching them with advanced cameras can provide a much clearer picture of drug safety than older methods. By capturing the subtle, early changes in cell structure and accounting for the natural differences between people, this approach offers a powerful new way to protect patients from liver injury. It represents a shift from simply asking if a drug kills cells to asking how it changes them, providing a deeper, more human-relevant understanding of how medicines interact with our bodies. As drug development continues to move away from animal testing, tools like this will be essential for ensuring that new therapies are both effective and safe for the diverse population they are meant to serve.