Nature-inspired method helps study protein clumps in neurodegenerative diseases

Neurodegenerative conditions such as Alzheimer’s and Parkinson’s disease are characterized by the buildup of misfolded proteins in the brain. Understanding whether and how these protein clumps contribute to disease is key for developing effective treatments, but traditional methods for studying these aggregates rely on fluorescent tags that can alter the proteins’ natural behavior, which may lead to inaccurate results. Now, NCCR Bio-Inspired Materials researchers have developed a method that uses artificial intelligence (AI) to detect protein aggregates in living cells without the need for fluorescent labels.

The approach allows for accurate, real-time measurements of protein aggregates, providing a better way to investigate the biological processes underlying neurodegenerative diseases. “We wanted to develop a tool to study biological processes as accurately as fluorescence tools allow us, but in a label-free manner,” says NCCR PI and study senior author Aleksandra Radenovic, professor of biological engineering at the École Polytechnique Fédérale de Lausanne.

The team draw inspiration from nature, Radenovic says, emphasizing that living organisms don’t have artificial labels such as green fluorescent proteins. What’s more, the researchers used neural networks, which work similarly to the human brain by mimicking the way neurons interact and learn from experience. This approach allows the neural network to improve its accuracy over time.

Radenovic and her team used cells that overexpress a mutant version of a protein that leads to the formation of aggregates that are observed in Huntington’s disease — a neurodegenerative disorder that leads to movement, cognitive and psychiatric symptoms. Using a custom-built microscope, the team collected bright-field and fluorescence images of these cells.

Then, the researchers trained a neural network that could accurately identify these aggregates in 2D images without fluorescent tags. This label-free method is a powerful tool for studying protein aggregation in living cells, says study lead author Khalid Ibrahim, a PhD student in Radenovic’s lab. “Our approach has a 96% accuracy in detecting true aggregates, showing great promise for advancing research in neurodegenerative diseases,” he says.

The tool, which the researchers dubbed label-free identification of NDD-associated aggregates (LINA), was effective under varying conditions, including low light and different cell lines. LINA not only identified protein aggregates in living cells but could also track their growth over time, the researchers reported in Nature Communications.

Understanding the growth rate of these aggregates may help identify new treatment targets. In the future, the researchers hope to expand this technology for broader use in drug discovery, which may provide pharmaceutical companies with more reliable information compared to traditional fluorescent methods, Ibrahim says.

However, he adds, the current tool is limited to studying cells. The researchers aim to expand its use to various sample types, including tissues and organoids, which could boost its relevance in studying diseases.

Reference: Ibrahim, K. A.; Grußmayer, K. S.; Riguet, N.; Feletti, L.; Lashuel, H. A.; Radenovic, A. Label-Free Identification of Protein Aggregates Using Deep Learning. Nat Commun 2023, 14 (1), 7816. https://doi.org/10.1038/s41467-023-43440-7.