Skip to main content
Premium Trial:

Request an Annual Quote

New Protein Design Methods Promise to Boost Nanopore Applications

Premium

NEW YORK – Two teams led by researchers from the University of Washington and the VIB-VUB Center for Structural Biology in Belgium have developed new de novo protein design methods that could be harnessed to protein nanopores for sensing and sequencing applications. 

Both teams published their methods in Science last month. The Belgian study illustrated that artificially synthesized beta-barrel proteins can be tapped as transmembrane nanopores and have properties of biological sensors, while the US study described a workflow to design binder proteins that attach to target small molecules with high affinity.

"You can design a protein that can specifically bind to the specific metabolite that you are interested in and then [develop a] nanopore that is providing the readout," said Anastassia Vorobieva, a group leader at the VIB-VUB Center for Structural Biology and the corresponding author of the Belgian study. "I think this is where the two studies come nicely together."

Currently, nanopore sensor development primarily relies on engineering naturally occurring protein channels, providing researchers with only a limited number of options and suboptimal starting points, Vorobieva and her coauthors wrote in their study. In contrast, de novo protein design can, in theory, generate an unlimited number of new pores with any desired properties.

The team decided to develop beta-barrel proteins, which are made of beta-strands, from scratch and achieve the desired transmembrane scaffolds that allow the flow of target molecules through the pore while generating distinguishable signals. 

According to Vorobieva, who previously worked as a postdoctoral researcher in the lab of UW professor David Baker, a corresponding author of both studies, a major challenge in designing beta-barrel nanopores is the lack of computational models to help predict protein folding in lipid membranes, given that most existing tools in the field are developed based on protein folding dynamics in water. 

"The major problem we are trying to solve is to understand what is fundamentally necessary for a computational model to make a membrane protein that is able to fold in real life," she said, adding that her team developed its own model using existing knowledge of membrane proteins and their physical properties.

For their study, Vorobieva and collaborators built a series of transmembrane beta-barrel nanopores with varying diameters and shapes by manipulating the placement of glycine residues, where beta-strand bending takes place.

After experimentally confirming that the nanopore structures corresponded with the design models, using nuclear magnetic resonance and X-ray crystallography, the researchers further tested the artificial pores in planar membranes from diluted detergent solution and showed that they had stable and distinct conductance at positive and negative voltage.

Overall, the ability to custom-design transmembrane nanopores will open the door for potential sensing and sequencing applications, Vorobieva noted.

To that end, Linna An, a postdoctoral researcher in the Baker lab, developed a pipeline that can help design high-affinity protein binders for specific small molecules, and fused binder proteins with de novo-synthesized nanopores to form an artificial ligand-gated sensor.

According to An, the lead author of the UW-led paper, an important component of the study was the development of a deep learning-based pipeline that can analyze small molecule-protein interactions and generate pseudocyclic peptides with high binding affinity to target analytes of interest. 

"Our method deals with small molecule-protein interfaces [and informs] you how to manipulate those instead of having to go to the wet lab and screen thousands of [potential protein binders]," said An. "The idea is, if you can find out some small molecule binding protein [computationally] without using all those expensive screening [methods], that will be a really good method."

For their study, An and colleagues docked small molecules of interest to deep learning–generated peptide pseudocycles, which have varying shapes due to the geometry and number of repeat units surrounding the central binding cavity. From there, they tested better-performing binders experimentally, and the design of the best hits were "extensively resampled," the authors noted.

Currently, the experimental success rate for the deep-learning pipeline is around 0.1 percent, An said, meaning out of 1,000 candidate binders generated by the method, there is likely one to show high affinity to the target small molecule. "We have a good starting point," she noted, adding that the team is working to improve this to 1 percent. Beyond small molecule-protein interactions, An said, the team plans to apply the pipeline to enzyme design and manipulation.

Meanwhile, Vorobieva said the computational pipeline developed for the de novo nanopore synthesis can still be "very complex use," and the goal for the team is to continue streaming the method and improve its usability. Additionally, she said the team plans to further engineer the artificial nanopores to be able to detect multiple targets at once.

The new methods "increase the toolbox that you might have for a particular [nanopore] application, not just limited to what might be in nature," said Jeff Nivala, a molecular engineering professor at UW who was not involved in either study.

According to Nivala, one potential advantage of developing protein nanopores from scratch is the ability to design and sculpt those molecules to achieve desired characteristics, alleviating researchers from the constraint of natural nanopores. Additionally, he applauded the authors' efforts to develop pipelines that can help predict protein folding and small molecule interactions, which he said can "greatly reduce" the search space.

Despite these promises, Nivala thinks that designing nanopores de novo "is not necessarily an easy process" at this point, given that it still requires significant experimental optimization due to the current insufficient understanding of nanopore biology.

"To some degree, there is still going to be some empiricism with the process, because we are still understanding what makes certain pores better than others," he explained.

"I don't think anyone really knows what we need to design for — that's the big challenge," Vorobieva agreed. "We don't really know what property of the nanopore would make it perfect [for specific applications]."

The ability to synthesize artificial nanopores with varying characteristics will also hopefully allow researchers to study them, she added, further propelling the understanding of pore function and biology.

Nevertheless, Nivala believes there is still a need for the field study and improve pores occurring in nature. "​​I don't think we have gotten to the point in protein design that we can say that we can do better than nature," he said. "I think there is still a lot of utility in not only looking at what nature has evolved but also learning from nature."