Many drug and antibody discovery pathways focus on intricately folded cell membrane proteins: when molecules of a drug candidate bind to these proteins, like a key fitting into a lock, they trigger chemical cascades that alter cellular behavior. But because these proteins are embedded in the lipid-containing outer layer of cells, they are difficult to access and insoluble in water-based solutions (hydrophobic), making them difficult to study.
“We wanted to get these proteins out of the cell membrane, so we redesigned them as soluble, hyperstable analogues, which look like membrane proteins but are much easier to work with,” explains Casper Goverde, a doctoral student in the Protein Design Laboratory. and Immunoengineering (LPDI) at the School of Engineering.
In a nutshell, Goverde and a research team at LPDI, led by Bruno Correia, used deep learning to design soluble synthetic versions of cell membrane proteins commonly used in pharmaceutical research. While traditional detection methods rely on indirectly observing cellular reactions to drug candidates and antibodies, or painstakingly extracting small amounts of membrane proteins from mammalian cells, the researchers’ computational approach allows them to eliminate cells of the equation. After designing a soluble protein analog using their deep learning process, they can use bacteria to produce the modified protein in bulk. These proteins can then bind directly in solution to molecular candidates of interest.
“We estimate that producing a batch of soluble protein analogues using E.coli is approximately 10 times cheaper than using mammalian cells,” adds doctoral student Nicolas Goldbach.
The team’s research was recently published in the journal. Nature.
Flipping the script on protein design
In recent years, scientists have successfully leveraged artificial intelligence networks that use deep learning to design new protein structures, for example by predicting them based on an input sequence of amino acid building blocks. But for this study, the researchers were interested in protein folds that already exist in nature; what they needed was a more accessible and soluble version of these proteins.
“We had the idea of reversing this deep learning process that predicts the structure of proteins: if we input a structure, can it tell us the corresponding amino acid sequence?” explains Goverde.
To achieve this, the team used Google DeepMind’s AlphaFold2 structure prediction platform to produce amino acid sequences for soluble versions of several key cell membrane proteins, based on their 3D structure. They then used a second deep learning network, ProteinMPNN, to optimize those functional soluble protein sequences. The researchers were pleased to find that their approach showed remarkable success and precision in producing soluble proteins that maintained parts of their native functionality, even when applied to highly complex folds that have so far eluded other design methods.
“The holy grail of biochemistry”
A particular triumph of the study was the project’s success in designing a soluble analog of a form of protein known as a G protein-coupled receptor (GPCR), which accounts for about 40% of human cell membrane proteins and is a important pharmaceutical target.
“We showed for the first time that we can redesign the shape of the GPCR as a stable soluble analog. This has been a long-standing problem in biochemistry, because if you can make it soluble, you can detect new drugs much faster and more easily,” says the LPDI scientist Martín Pacesa.
The researchers also see these results as proof of concept for the application of their project to vaccine research and even cancer therapy. For example, they designed a soluble analog of a type of protein called claudin, which plays a role in making tumors resistant to the immune system and chemotherapy. In their experiments, the team’s soluble claudin analog retained its biological properties, reinforcing the project’s promise of generating interesting targets for pharmaceutical development.