A fluorescent micrograph of cancer cells. A new image-processing algorithm helps to unmix spatially overlapping proteins in such images.Credit: Nicola Ferrari/Getty Images
A novel technique for unmixing the signals from as many as 15 spatially overlapping fluorophores is allowing the identification and separation of spatially overlapping proteins in heterogenous biological samples using just one imaging round (J. Seo et al. Nature Commun. 13, 2475; 2022). This approach promises to accelerate biological research and help with the early detection and diagnosis of cancer.
Fluorescence microscopy is one of the main workhorses of research in the life sciences and medicine. It allows researchers to stain and visualize specific molecules in a sample and to discover the locations and movements of different components in cells.
In general multicolour-experiments, only four or five different fluorophores with distinct excitation and emission spectra can be used simultaneously in a sample. However, the increasing complexity of biological and medical research questions demands the utilization of more dyes per sample.
Addressing this gap, eight researchers in South Korea have developed a sophisticated technique that makes it possible to ‘unmix’ multiple dyes coupled to spatially overlapping proteins in a biological sample without using reference spectra for each fluorophore.
The team demonstrated their technique, which they named PICASSO, by imaging samples from the mouse brain. They stained tissue samples with multiple antibodies that target proteins that are strongly expressed and have well-known expression patterns. The images obtained from various brain regions were clearly unmixed using PICASSO.
“Under a wide range of experimental conditions, PICASSO was able to accurately separate the mixed images into individual images, each containing the signal of just one protein,” says Jae-Byum Chang of the Korea Advanced Institute of Science and Technology in Daejeon, South Korea. “PICASSO is especially advantageous in cases where it is difficult to measure the exact emission spectra of fluorophores, or when these spectra vary over space.”
“PICASSO operates on the assumption that the amount of shared information between two images increases as the images become more mixed,” explains Chang. “Based on this assumption, PICASSO gradually reduces the amount of mutual information shared by images acquired in different spectral ranges.”
Importantly, PICASSO is a ‘blind’ unmixing technique, meaning that it doesn’t require reference spectra — that is, those taken from a region of an image containing just one dye. This simplifies the multiplexed imaging process. “Measuring precise emission spectra can sometimes be challenging and time consuming for complex specimens,” says Chang. “For such cases, a blind unmixing approach such as PICASSO can be a powerful alternative.”
The researchers are delighted by the significant surge in interest in multiplexed imaging today. In the past five years, numerous multiplexed imaging techniques, including their own, have emerged. “Some of these innovative techniques have already been commercialized and are actively being used in laboratories,” says Chang. “We’re excited to be contributing to this rapidly expanding field.”
The team intends to take their research further. “PICASSO represents our first step towards developing more-robust blind unmixing algorithms that can effectively handle images with high levels of noise or low signal intensities,” says Chang.