UCLA Study Uses AI to Decode Tumor Mitochondria

In a new study published in Nature, UCLA researchers used a combination of positron emission tomography (PET) and electron microscopy to generate 3-dimensional ultra-resolution maps of mitochondrial networks in lung tumors of genetically engineered mice. Mitochondria, also known as the “powerhouses” of cells, have been found to play a critical role in cancer cell metabolism and energy production. However, until now, little was known about the relationship between mitochondria’s structural organization of networks and their functional bioenergetic activity at the level of whole tumors. 

Using an AI deep learning technique, the researchers were able to categorize tumors based on mitochondrial activity and other factors, quantifying the mitochondrial architecture across hundreds of cells and thousands of mitochondria within the tumor.

The researchers examined two main subtypes of non-small cell lung cancer (NSCLC) — adenocarcinomas and squamous-cell carcinomas — and found distinct subpopulations of mitochondrial networks within these tumors. Notably, they discovered that these mitochondria frequently organized themselves with organelles to create unique subcellular structures that support tumor cell metabolism and mitochondrial activity.

The team, led by Mingqi Han, Ph.D., anticipates that mitochondrial populations in human cancer samples will not be mutually exclusive to their respective tumor subtype, but instead there will be a spectrum of activity.

The investigators say these findings provide key information about motochondria function in cancer cells, potentially leading to new approaches for cancer treatment. By understanding how mitochondrial networks are structurally and functionally regulated in non-small cell lung cancer at an in vivo level, researchers may be able to develop new treatments that target tumor-specific nutrient preferences.