After proteins are synthesized in a cell, they don’t just sit there. They get straight to work—alone or in combination—performing nearly every task in a cell’s life, moving from nucleus to cytoplasm to membrane. Cell biologists know this happens, and they’ve done a decent job of describing a multitude of these activities and interactions. But of course, we always want to know more, especially when it comes to what happens in a living cell.
Manuel Leonetti, group leader in Quantitative Cell Science at the Chan Zuckerberg Biohub, a nonprofit research institute associated with the University of California, San Francisco, UC Berkeley, and Stanford University, is taking the characterization of “proteins in the wild” a step further with his project called OpenCell.
“OpenCell is an open-source project designed to map the intracellular organization of the proteome,” Leonetti said. “We want to define where a protein is localized within the cell and what molecular interactions that protein makes with other proteins,” Leonetti added. “If we can map these two things for every protein, we could have a full description of how the cell is wired internally.”
To accomplish this, the Leonetti lab deconstructs the cell one protein at a time. Each protein is tagged using CRISPR with a fluorescent reporter (such as green fluorescent protein). Then this fluorescently tagged cell line is used in live-cell imaging experiments to map where the protein localizes. Through advanced imaging techniques, machine learning, data science, and software engineering, OpenCell creates a snapshot of different proteins at steady state, measuring their quantity and mapping their position in human embryonic kidney cells (HEK293T, a widely used model cell line). To further chart what the proteins do, the interactions between the tagged protein and other proteins in the cell are measured using immunoprecipitation and mass spectrometry, achieved through a partnership with Matthias Mann’s lab at the Max Planck Institute of Biochemistry in Germany. The OpenCell dataset was described in a recent article (https://www.science.org/doi/10.1126/science.abi6983) and an associated pre-print (https://www.biorxiv.org/content/10.1101/2021.03.29.437450v2).
Leonetti credits OpenCell with being built on the foundation of work previously done in yeast (Saccharomyces cerevisiae), specifically the extensive yeast GFP library built by Erin O’Shea and Jonathan Weissman. “It illustrated that endogenous GFP tagging could illuminate the proteome at a very general level,” he said.
So far, for human cells, Leonetti says, his group has mapped 1,310 different proteins, which might sound like a lot, but this represents only 7% of the human genome. If you go to the OpenCell website (https://opencell.czbiohub.org/), select protein targets by name, view a microscopy gallery, or search for proteins typically found in various subcellular locations, such as the endoplasmic reticulum or the mitochondria. These images and data sets are downloadable, and anyone can use them in their research. He also said that educators could use them to develop lectures, projects, and experiments. In the future, Leonetti said that OpenCell would be switching away from the HEK293T cell lines to induced pluripotent stem cell (iPSC) lines, which will help researchers better understand the activity of proteins in different types of human tissue. The lab is also turning to light-sheet microscopy to image protein localization through time, in particular during cell division.
Working on OpenCell is not without challenge, Leonetti explained. One of the limiting factors is that many endogenous proteins are not expressed at easily detectible levels, despite working with fluorescent tags that are as bright as currently available, such as mNeonGreen. “Fifty percent of the proteins in the cell are expressed at a level that is a little bit too low for us to be able to look at under a microscope,” Leonetti said.
Even so, Leonetti said that he has been surprised at how much they could learn from these steady state images. Working with CZ Biohub colleague Loïc Royer, they built a machine learning algorithm to encode the proteins’ spatial distributions quantitatively to compare the differences and similarities between two proteins. “We discovered that no two proteins look the same,” he said. “The localization of a protein in the cell is extremely specific. And it’s only when two proteins are part of the same pathway or protein complex that the localizations start to look alike. Our data show that there is a degree of micro-compartmentalization in the human cell that is probably higher than what I would have expected.” He added that the team could predict future protein—protein interactions simply by comparing fluorescent images.
“The cell and the specificity of how the cell is organized keeps surprising us!” Leonetti said.
If you want to hear the full interview with Manuel Leonetti, listen to the ASCB Pathways podcast here: https://anchor.fm/dashboard/episode/e1ita8r
About the Author:
Mary Spiro is ASCB's Strategic Communications Manager.