Imaging 45 types of biomolecules

Researchers at KAIST (Profs. Jae-Byum Chang and Young-Gyu Yoon) have developed a microscopy technique named PICASSO that expands the number of simultaneously available fluorescent molecules from 4 to 15, enabling 45-color imaging of 45 biomolecules with three rounds of imaging.

Article | Fall 2022

How can we take an image of ten or even more biomolecules in a biological tissue sample, by expanding the color palette we can use for imaging, that contains an enormous amount of information which may tell us how to treat a certain type of cancer?

Each color corresponds to a specific wavelength in a continuous spectrum, which means the number of colors is simply infinite. Thus, the amount of information that can be carried in a colored image, in theory, is nearly unlimited. This suggests that it may be possible to capture a huge amount of information through the multi-color imaging of a biological tissue sample. Nearly 80 years ago, researchers developed immunostaining (i.e., immune plus staining), a technique that uses an antibody — derived from animals’ immune systems — whose one arm can bind a specific molecule while the other arm binds to a fluorescent light-emitting molecule. This allows us to selectively stain a specific biomolecule using a fluorescent molecule that has a specific color. This opened the possibility of the simultaneous staining and imaging of various biomolecules to reveal their spatial distributions in tissue.

Unfortunately, the number of fluorescent molecules we can simultaneously use for imaging has remained limited to four, as the light emitted by a fluorescent molecule has a wide spectrum which makes it difficult to distinguish the light from different fluorescent molecules. Due to such spectral overlapping between different fluorescent molecules, an image acquired by detecting a spectral range is manifested as linear mixtures of the images of different biomolecules.

Researchers at KAIST (led by Profs. Jae-Byum Chang and Young-Gyu Yoon) have developed a microscopy technique, named PICASSO, to tackle this problem by applying principles from information theory. First, images that contain the linear mixtures are taken. Importantly, each image is acquired using a unique spectral detection range so that it has a unique “mixing ratio.” At this stage, the images look very similar to one another, except that different biomolecules appear slightly brighter in different images. Thereafter, the images are fed into an iterative algorithm that ‘unmixes’ the mixtures by minimizing the mutual information (i.e., the amount of shared information) between each pair of images.

Through such algorithmic unmixing and three rounds of imaging, PICASSO enables 45-color imaging of spatially and spectrally overlapping fluorescent molecules —revealing the spatial distribution of 45 types of proteins — without compromising on the quality. This will help us disclose the cellular heterogeneities of tumor microenvironments, especially with regard to heterogeneous populations of immune cells, which are closely related to cancer prognoses and the efficacy of cancer therapies.