David Baker’s research team at the Institute for Protein Design has unveiled a new technique for sequestering membrane proteins (Science 2026, DOI: 10.1126/science.adr3817).
Proteins bound in cell membranes are not soluble in water or stable outside that amphiphilic membrane. This makes working with them difficult, whether to determine their structure or to potentially use them as a vaccine antigen.
The common approach is to use a detergent to detach such proteins from the membrane, but this method poses several problems: the detergent needs to stay with the protein of interest, and in the case of potential vaccines, the detergent is a contaminant, says Ljubica Mihaljević, one of the team members behind the new work.
While it is also possible to mutate some proteins enough to make them water soluble, doing so sacrifices studying the protein’s true structure.
Mihaljević and her colleagues found a way around the issue with water-soluble RFdiffused amphipathic proteins, or WRAPs, that can bind to the membrane proteins and keep them stable and water soluble without the need for detergent. RFdiffused means that these proteins are designed from scratch with artificial intelligence. WRAPs are designed to form a barrier around the hydrophobic regions of the membrane protein, stabilizing it and keeping it soluble in water.
For membrane proteins that researchers want to make in Escherichia coli, the team designs a protein, orders a synthetic DNA sequence, and instructs the bacteria to express the corresponding WRAP along with the membrane protein of interest. Mihaljević says these proteins, once translated, never get to the membrane, with the WRAPs sequestering them instead.
The team used this technique on several membrane proteins in Treponema pallidum, the bacteria responsible for syphilis. The WRAP-bound versions of these proteins could be deployed as possible antigens for a syphilis vaccine.
“I’m excited to release the code to the general public for people who have their favorite membrane protein and know the structure, or it can be predicted with AlphaFold, to use WRAPs for whatever application they like,” Mihaljević says. “That’s the best way to show that your science is working.”
