Understanding how proteins interact inside living cells is fundamental to biology. Yet, visualizing these interactions with high precision—and at the scale and condition at which they actually occur—has remained a major technical challenge. In a new study in Proceedings of the National Academy of Science (PNAS), NSF-Simons Postdoctoral Fellow Xingchi Yan and colleagues in the lab of MCB faculty member Maxim Prigozhin introduce a novel method that quantifies protein-protein interactions at nanoscale resolution in cells. “Protein-protein interactions are central to pretty much everything that cells do,” says Prigozhin. “But determining whether two proteins interact or not inside a cell is a very challenging task. With this paper, Xingchi, who is a very talented mathematician who joined us for a year to work on important problems at the interface of mathematics and biology, achieved this goal.”
The paper introduces a computational framework that, in combination with advanced single-molecule imaging, allows researchers to directly estimate the functional binding from nanoscale mapping of cellular proteins. “This is a very fundamental question,” Yan says. “Biomolecular interactions govern key cellular processes, such as gene regulation, cell signaling and enzymatic catalysis. When labeling and imaging multiple biomolecules—such as protein-protein or protein-RNA—within a cell in situ at nanoscale resolution, how can true molecular interactions be distinguished from random colocalization due to spatial proximity? That’s not easy to answer, especially at the high densities we see in real, physiological conditions.”
At the heart of the new method is single-molecule localization microscopy (SMLM), a technique that allows scientists to pinpoint the positions of individual molecules with extreme precision—down to a few nanometers. While SMLM has been around for years, using it to quantify how many molecules are actually interacting (rather than just co-localized) within specific subcellular compartments has been a major hurdle. High protein density, stochastic molecular diffusion, and variation in membrane organization all complicate analysis.
“The difficulty is that proteins are far from uniform—they diffuse, they bind and dissociate, and they’re differentially packed compartment-wise,” Yan says. “We wanted to create a method that works in this heterogeneous environment, one that better reflects the true biological conditions.”
The team’s mathematical algorithm analyzes SMLM images via calculating proximity probabilities between all possible interactions and identifying pairings through graphical methods. It then iteratively corrects for random colocalization using Monte Carlo methods to estimate the amount of true protein-protein interactions for both steady and transient cellular processes. The method was carefully validated with simulations and experimental data, showing that it significantly improves upon existing approaches. The broader vision of Prigozhin’s lab is to use this new methodology in tandem with time-resolved nanoscale bioimaging techniques—also under development in the group—to capture how protein interactions evolve in response to drugs or stimuli.
“Max’s lab is working on building exciting new instruments to take images of multiple proteins at fast and precise time points after introducing a ligand,” Yan says. “By combining that with our method, we’ll be able to directly measure temporal and spatially-dependent reaction kinetics in cells. It’s incredibly exciting to collaborate with the team, as the new measurement tools developed in the lab will unlock a whole range of new mathematical and computing challenges.”
Prigozhin sees this work as an essential first step toward a much larger goal. “It’s really a foundational project,” he says. “Not only did Xingchi devise a clever mathematical approach to identify protein-protein interactions at nanometer resolution, but the method performs very favorably compared to alternatives. I expect it will be used widely in the future by many labs that use high-resolution microscopy to image single proteins in situ.”
Yan, who joined the Prigozhin lab as an NSF-Simons Postdoctoral Fellow, brought an unusual skill set to the project. The NSF-Simons Center is a part of the Quantitative Biology Initiative at Harvard, and its major focus is to train early-career researchers who will emerge as leaders in the field. “Xingchi is passionate about addressing important problems at the intersection of applied physics, mathematics, and biology,” says Prigozhin. “I’m very fortunate to have recruited Xingchi—he has been fantastic: super creative, independent, and rigorous.”
The success of the project was also the result of close collaboration with other members of the team. “It’s important to highlight the contributions of Polly Yu and Arvind Srinivasan,” Prigozhin adds. Yu was an independent NSF-Simons Fellow and is now an Assistant Professor of Mathematics at the University of Illinois at Urbana-Champaign. She and Xingchi worked together on many algorithmic and computational aspects. Arvind, a PhD candidate in Prigozhin’s lab who will graduate this fall, provided most of the experimental data. “The work would not have been nearly as compelling without this experimental validation,” Prizgozhin says, adding, “This work could not have been done without the support of the NSF-Simons Center for Mathematical and Statistical Analysis of Biology at Harvard. It’s a wonderful example of how interdisciplinary training and support can lead to real breakthroughs.” Yan adds: “I’m especially grateful to Max, the incredible team, and the amazing QBIO and MCB communities for the support and the many insightful discussions that helped shape this work.”
