Cell biologists have long puzzled over the ability of cells to change shape, to move to different locations and then divide or differentiate into entirely different types. Much of their work has focused on the genes involved in these processes and the signaling pathways that control them.
One largely hidden factor in all this is the forces that cells exert on their surroundings as they change shape and push and pull themselves along. Indeed, cell biologists have begun to suspect that these forces play a significant role, not just in locomotion but in proliferation, differentiation and other processes. Indeed, the most recent thinking is that cells can sense the forces at work and then regulate their behavior accordingly.
This suspicion has led to a greater focus on these forces and an emerging science called mechanobiology. One of the enabling technologies behind mechanobiology is a technique called traction force microscopy, which measures the forces a cell exerts based on the wrinkles it creates on the surface it sits on. This in turn reveals the forces involved in processes such as cellular migration, differentiation and in disease. It also shows how these forces change when the cells are genetically modified or bathed in drugs.
There is a problem, however. Traction force microscopy is a multistage process that is time consuming to perform. This limits where it can be applied and the scale on which it can be done.
Now a team at Osaka University in Japan has trained a machine vision system to calculate the forces a cell exerts simply by looking at microscope images. This dramatically simplifies the process of force measurement and opens up the possibility of large-scale analyses of the forces exerted by different cell types under different conditions. The hope is that the science of mechanobiology could suddenly help to better understand a wider range of phenomena in areas such as development and in the study and treatment of disease.
First, some background. Traction force microscopy relies on a special flexible substrate containing tiny fluorescent microspheres that reveal when the surface is deformed. These microspheres and the way they are distributed are easily imaged with a microscope.
A key feature of this substrate is that it is flexible enough to wrinkle when cells exert a force on it. What's more, the material properties are well characterized so it is straightforward to calculate the force required to make a given wrinkle.
The substrate is then seeded with cells, which exert forces that cause the material to wrinkle. Some image manipulation software extracts the pattern of wrinkles and the way it changes from a series of images taken via a microscope.
The cells are then removed allowing the substrate to return to its unstressed state. Comparing the pattern of wrinkles to the unstressed pattern shows how far the material has been deformed.
Finally, researchers use this comparison to calculate the pattern of forces — the stress field — that the cells have exerted.
Of course, this is a time-consuming process, particularly the cell removal and unstressed image stage. And this limits the way the technique can be applied.
Now Hohghan Li and colleagues at Osaka University have found a way to dramatically speed up and simplify this process. They began by creating a database of microscope images along with the stress fields that were extracted from them using conventional traction force microscopy.
Next, they used this database to train a machine vision algorithm, called a generative adversarial network, to recognize the stress field associated with microscope images. Finally, they used this machine vision algorithm to predict the stress fields from microscope images it had not seen before. The new technique is quicker because it does away with the need to remove the cells and take a reference image of the substrate at the end of the process.
Visualizing Cellular Forces
The results are impressive. The researchers found that the algorithm could accurately determine the forces at work on a substrate purely from looking at the microscope image. "Our system provides a powerful tool to evaluate the cellular forces efficiently because the forces can be predicted just by observing the cells under the microscope," they say.
The team calls the new technique "wrinkle force microscopy" and expect great things from it. "Given that early stages of drug screening require testing a massive number of candidate compounds, our system with the potentially high-throughput data analysis capability will be useful particularly in such screening studies," they say.
That could reveal the role of mechanobiology and cellular forces in all kinds of diseases such as osteoporosis, fibrosis, heart failure and even cancer. And beyond that, the screening process that wrinkle force microscopy allows could help in the development of the next generation of drugs to treat these conditions.
Ref: Wrinkle Force Microscopy: A New Machine Learning-Based Approach to Predict Cell Mechanics from Images: arxiv.org/abs/2102.12069