Skip to main content
Premium Trial:

Request an Annual Quote

New Approach Could Improve Quantitative Analysis of Super-resolution Microscopy Protein Data


NEW YORK (GenomeWeb) – A team from Indiana University School of Medicine, Howard Hughes Medical Institute, and the University of California system has devised an approach to analyzing super-resolution microscopy data that allows better determination of the stoichiometry of protein complexes in living cells.

Detailed in a study published last week in Proceedings of the National Academy of Sciences, the method could add to SR microscopy's utility as a tool for studying molecular interactions in vivo, Steven Presse, an IUSM researcher and senior author on the paper, told GenomeWeb.

Capable of detecting proteins and other molecules at spatial resolutions as high as 20 to 30 nanometers, SR microscopy methods like photoactivated localization microscopy (PALM) can, in principle, enable researchers to generate in vivo quantitative data at the single molecule level. 

To date, however, the method has not fulfilled this promise, Presse said, due in significant part to the difficulty of confidently assigning detected signals to a single protein.

The PALM method is based on SR microscopy detection of fluorophores specific to a particular protein. These fluorophores are genetically encoded in the protein of interest and then, once expressed in the living cell, photoactivated using low light.

This results in stochastic activation of individual fluorophores such that a single fluorphore lights up at a time, which allows for temporal separation of proteins located too close together to be distinguished spatially.

"The goal is not to get them all activated at the same time, because you are trying to get very high spatial resolution," Presse said. "You want to separate them in time not in space, because they are all going to be bunched up in one small pixel of your camera."

The method's sticking point, however, is that the fluorophores don't typically just go on and off but, rather, "blink," meaning that a single protein can produce several bursts of signal. This phenomenon, Presse said, has limited PALM's usefulness as a quantitative method.

"If these things just turned on and off it would be really easy to count them," he noted. "But they don't do that, they blink. So you have to somehow be able to distinguish the flickering from what is actually an on–off event to determine how many [proteins] you have."

To tackle this problem, Presse and his colleagues looked to mathematical methods applied in the 1970s and 1980s to the study of ion channel gating, which, he said, presented a similar challenge.

In early ion channel gating experiments, researchers would "basically isolate one channel and monitor current fluctuations or stochastic changes between high and low voltage," he said. "And typically the goal of ion channel data analysis is to be able to back out the kinetics of opening and closing of the channel."

Experimentally, researchers could observe two ion channel states — open and closed — Presse said.

"But there could potentially be multiple open or multiple closed states," he noted. "So the idea is that what you see, which is either conducting or non-conducting, is less than the full range of options. The underlying kinetics are much more complicated than the simple binary output that you see."

The PALM blinking problem displays similar kinetics, where the observable "bright" and "dark" states don't fully account for the complexity of the data.

"When something is bright, we don't know if it just turned on or if this is the third time it flickered," Presse said. "So bright can mean many things and dark can mean many things, and the idea is to use the math to tease out rates of flickering that will allow us to correct for the flickering and ultimately determine how many protein subunits make up a protein complex."

Specifically, the authors used a version of continuous time aggregated Markov model methods, which, they said, allowed them to accurately quantify proteins in simulated data as well as determine the kinetics behind the fluorophore excitation to further improve the method's accuracy.

The ability to generate quantitative data from the PALM analysis is significant for studying protein complexes in that PALM provides significantly higher resolution than most other in vivo methods, Presse said.

"If you don't have good spatial resolution then you might have a bunch of protein complexes all next to one another, and you'll never quite be able to resolve one from another, and so you won't be able to determine the stoichiometry," he said.

While PALM experiments have typically looked at one type of protein at a time, he and his colleagues' method could be applied to multiplexed analyses, as well, Presse added, noting that the math "is not conceptually more difficult."

The approach presented in the group's PNAS paper is not the first attempt to address the PALM "blinking" problem. As the authors noted, researchers have previously tried to work around the issue by establishing a time threshold within which emission bursts are considered to come from a single molecule and outside of which they are considered to have come from different molecules.

Such "thresholding" approaches, however, require advance knowledge of fluorophore kinetic rates in order to best determine the threshold, which limits their accuracy and flexibility, the PNAS authors said.